TW202211668A

TW202211668A - Shooting processing method, electronic equipment, and computer-readable storage medium

Info

Publication number: TW202211668A
Application number: TW110122987A
Authority: TW
Inventors: 白煥鵬; 王超; 王美榮
Original assignee: 大陸商上海商湯智能科技有限公司
Priority date: 2020-09-14
Filing date: 2021-06-23
Publication date: 2022-03-16
Also published as: WO2022052475A1; CN112165573B; CN112165573A

Abstract

The embodiments of the present disclosure disclose a shooting processing method, electronic equipment, and computer-readable storage media, the shooting processing method includes: acquiring current brightness information of the environment in which the shooting device is located, wherein the shooting device includes at least two image acquisition components; At least one image acquisition component is adjusted to a shooting mode matching the current brightness information; wherein, different shooting modes use different lights to form images.

Description

拍攝處理方法和電子設備、電腦可讀儲存介質Shooting processing method and electronic device, and computer-readable storage medium

本發明關於圖像檢測技術領域，特別是關於一種拍攝處理方法和電子設備、電腦可讀儲存介質。The present invention relates to the technical field of image detection, and in particular, to a photographing processing method, an electronic device, and a computer-readable storage medium.

目前圖像拍攝已廣泛應用在各領域中，例如監控、門禁控制、人臉驗證等領域。通常，需要採用拍攝設備對目標進行拍攝，進而再利用拍攝得到的圖像進行如目標識別、活體檢測等處理。At present, image capture has been widely used in various fields, such as monitoring, access control, face verification and other fields. Usually, it is necessary to use a photographing device to photograph the target, and then use the photographed image to perform processing such as target recognition and living body detection.

一般，傳統場景中使用都是傳統的拍攝設備，通常採用的是單攝影頭拍照圖像採集到的圖像。但是單攝影頭由於本身特性的限制，致使成像的品質一直得不到提高。Generally, traditional shooting equipment is used in traditional scenes, and images captured by a single camera are usually used. However, due to the limitation of its own characteristics, the imaging quality of the single camera has not been improved.

本發明實施例至少提供一種拍攝處理方法和電子設備、電腦可讀儲存介質。The embodiments of the present invention provide at least a photographing processing method, an electronic device, and a computer-readable storage medium.

本發明實施例第一方面提供了一種拍攝處理方法，包括：獲取拍攝設備所處環境的當前亮度資訊，其中，拍攝設備包括至少兩個圖像採集組件；將至少一個圖像採集組件調整為與當前亮度資訊匹配的拍攝模式；其中，不同拍攝模式是採用不同的光形成圖像。A first aspect of the embodiments of the present invention provides a photographing processing method, including: acquiring current brightness information of an environment where a photographing device is located, wherein the photographing device includes at least two image acquisition components; and adjusting the at least one image acquisition component to match the Shooting modes that match the current brightness information; wherein, different shooting modes use different lights to form images.

其中，將至少一個圖像採集組件調整為與當前亮度資訊匹配的拍攝模式，包括：在當前亮度資訊大於亮度閾值的情況下，將至少一個圖像採集組件調整為第一拍攝模式；在當前亮度資訊不大於亮度閾值的情況下，將至少一個圖像採集組件調整為第二拍攝模式。因此，利用亮度資訊是否大於閾值的條件來調整圖像採集組件的拍攝模式，使得獲取到的圖像的品質更好。Wherein, adjusting the at least one image acquisition component to a shooting mode matching the current brightness information includes: when the current brightness information is greater than the brightness threshold, adjusting the at least one image acquisition component to the first shooting mode; When the information is not greater than the brightness threshold, the at least one image acquisition component is adjusted to the second shooting mode. Therefore, the shooting mode of the image acquisition component is adjusted according to the condition of whether the brightness information is greater than the threshold value, so that the quality of the acquired image is better.

其中，第一拍攝模式為彩色拍攝模式，第二拍攝模式為紅外拍攝模式。因此，通過在亮度資訊大於閾值時，選擇彩色拍攝模式，當亮度資訊小於閾值時，選擇紅外拍攝模式，使得白天亮度較高紅外線過強時，減少紅外成像會受到強紅外線的干擾，而在夜晚亮度較暗時，選擇紅外拍攝模式，減輕了對可見光的依賴，從而提高了成像品質。The first shooting mode is a color shooting mode, and the second shooting mode is an infrared shooting mode. Therefore, when the brightness information is greater than the threshold, the color shooting mode is selected, and when the brightness information is less than the threshold, the infrared shooting mode is selected, so that when the brightness is higher during the day and the infrared rays are too strong, the infrared imaging will be reduced by the interference of strong infrared rays, and at night When the brightness is low, the infrared shooting mode is selected to reduce the dependence on visible light, thereby improving the image quality.

其中，拍攝設備中的每個圖像採集組件的拍攝模式相同。因此，當亮度資訊大於亮度閾值時，將兩個圖形採集組件皆調整為第一拍攝模式，當亮度資訊小於或等於亮度閾值時，則將兩個圖像採集組件皆調整為第二拍攝模式。Wherein, the shooting mode of each image acquisition component in the shooting device is the same. Therefore, when the brightness information is greater than the brightness threshold, both image capture components are adjusted to the first shooting mode, and when the brightness information is less than or equal to the brightness threshold, both image capture components are adjusted to the second shooting mode.

其中，在第一拍攝模式為彩色拍攝模式，第二拍攝模式為紅外拍攝模式的情況下，將至少一個圖像採集組件調整為第一拍攝模式，包括以下至少一個步驟：關閉拍攝設備的紅外補光燈；將拍攝設備的紅外截止濾光片移動至需調整為第一拍攝模式的圖像採集組件的入光通道中；將至少一個圖像採集組件調整為第二拍攝模式，包括以下至少一個步驟：打開拍攝設備的紅外補光燈；將拍攝設備的紅外截止濾光片移動至需調整為第二拍攝模式的圖像採集組件的入光通道以外。因此，在圖像採集組件為彩色拍攝模式時，說明環境亮度較高，紅外光比較強，因此，無需打開紅外補光燈，且將紅外截止濾光片放置在圖像採集組件的入光通道內，對紅外光進行過濾，減輕了紅外光對拍攝到的圖像的影響，而在圖像採集組件為紅外拍攝模式時，說明環境亮度較暗，才利用打開紅外補光燈，使得拍攝到的圖像更清楚即品質更好。Wherein, when the first shooting mode is the color shooting mode and the second shooting mode is the infrared shooting mode, adjusting at least one image acquisition component to the first shooting mode includes at least one of the following steps: turning off the infrared compensation of the shooting device. light; moving the infrared cut-off filter of the photographing device to the light incident channel of the image acquisition component that needs to be adjusted to the first photographing mode; adjusting at least one image acquisition component to the second photographing mode, including at least one of the following Steps: turn on the infrared fill light of the photographing device; move the infrared cut-off filter of the photographing device to outside the light incident channel of the image acquisition component that needs to be adjusted to the second photographing mode. Therefore, when the image capture component is in the color shooting mode, it means that the ambient brightness is high and the infrared light is relatively strong. Therefore, there is no need to turn on the infrared fill light, and the infrared cut-off filter should be placed in the light incident channel of the image capture component. Inside, the infrared light is filtered to reduce the influence of infrared light on the captured image. When the image acquisition component is in the infrared shooting mode, it means that the ambient brightness is dark, and the infrared fill light is turned on to make the captured The image is clearer and the quality is better.

其中，方法還包括：獲取至少兩幀初始圖像，其中，至少兩幀初始圖像為至少圖像採集組件對所處環境中的目標對象分別拍攝得到的；基於每幀初始圖像，對應得到包含目標對象的待檢測圖像；對至少兩幀待檢測圖像進行活體檢測，得到關於目標對象的活體檢測結果。因此，通過利用拍攝設備的至少兩個圖像採集組件拍攝的圖像而獲取到的待檢測圖像用於活體檢測相對於單攝影頭採集到的圖像進行活體檢測，前者在活體檢測過程中可以利用兩張圖像之間的視場差，從而提高活體檢測的準確性。Wherein, the method further includes: acquiring at least two frames of initial images, wherein the at least two frames of initial images are obtained by photographing the target object in the environment at least by the image acquisition component; based on each frame of the initial image, correspondingly obtained The to-be-detected images of the target object are included; in vivo detection is performed on at least two frames of the to-be-detected images to obtain the in vivo detection results of the target object. Therefore, the to-be-detected image acquired by using images captured by at least two image acquisition components of the photographing device is used for living body detection. Compared with the image collected by a single camera, the former is used for living body detection during the living body detection process. The difference in the field of view between the two images can be exploited to improve the accuracy of liveness detection.

其中，至少兩個圖像採集組件包括第一圖像採集組件和第二圖像採集組件；基於每幀初始圖像，對應得到包含目標對象的待檢測圖像，包括：從第一圖像採集組件採集到的若干幀第一初始圖像中，選出最終第一初始圖像；獲取第二圖像採集組件採集的與最終第一初始圖像對應幀的第二初始圖像；分別利用最終第一初始圖像和第二初始圖像，對應得到包含目標對象的兩幀待檢測圖像。因此，通過先在第一圖像採集組件採集到的若干張第一初始圖像，然後再利用選擇出的第一初始圖像以及對應的第二初始圖像獲得待檢測圖像，加快了選擇的速率，在一定程度上提高了獲取待檢測圖像的效率，而且相比於對兩個圖像採集組件的初始圖像分別進行選擇，本方案只需對其中一個圖像採集組件的初始圖像進行選擇，可以減少對處理資源的使用。Wherein, the at least two image acquisition components include a first image acquisition component and a second image acquisition component; based on each frame of the initial image, correspondingly obtaining an image to be detected containing the target object, including: acquiring an image from the first image From the several frames of the first initial images collected by the component, select the final first initial image; obtain the second initial image of the frame corresponding to the final first initial image collected by the second image acquisition component; use the final first initial image respectively; An initial image and a second initial image correspond to two frames of images to be detected that contain the target object. Therefore, by first acquiring several first initial images collected by the first image acquisition component, and then using the selected first initial image and the corresponding second initial image to obtain the image to be detected, the selection process is accelerated. This scheme improves the efficiency of acquiring the image to be detected to a certain extent, and compared with the selection of the initial images of the two image acquisition components, this scheme only needs to select the initial image of one of the image acquisition components. Like selection, you can reduce the use of processing resources.

其中，從第一圖像採集組件採集到的若干幀第一初始圖像中，選出最終第一初始圖像，包括：對每幀第一初始圖像進行目標檢測和跟蹤，得到每幀第一初始圖像所包含的第一目標對象；從若干幀第一初始圖像中，選擇第一目標對象符合活體檢測要求的第一初始圖像，作為最終第一初始圖像。因此，通過對若干張第一初始圖像進行目標檢測和跟蹤，選擇出符合條件的第一初始圖像，在一定程度上提升後續活體檢測的準確度。Wherein, selecting the final first initial image from several frames of the first initial images collected by the first image acquisition component includes: performing target detection and tracking on each frame of the first initial image to obtain the first initial image of each frame. The first target object included in the initial image; from several frames of the first initial images, select the first initial image of the first target object that meets the requirements of living body detection as the final first initial image. Therefore, by performing target detection and tracking on several first initial images, a first initial image that meets the conditions is selected, which improves the accuracy of subsequent living body detection to a certain extent.

其中，分別利用最終第一初始圖像和第二初始圖像，對應得到包含目標對象的兩幀待檢測圖像，包括：對第二初始圖像進行目標檢測，得到第二初始圖像所包含的第二目標對象；從最終第一初始圖像和第二初始圖像中，查找出匹配的一組第一目標對象和第二目標對象；利用最終第一初始圖像，得到包含查找出的第一目標對象的一幀待檢測圖像，以及利用第二初始圖像，得到包含查找出的第二目標對象的另一幀待檢測圖像。因此，對選擇出的第一初始圖像對應的第二初始圖像進行目標檢測，在一定程度上減輕了對系統資源的佔用。Wherein, using the final first initial image and the second initial image respectively to correspondingly obtain two frames of images to be detected containing the target object, including: performing target detection on the second initial image to obtain the content of the second initial image. The second target object of the A frame of an image to be detected of the first target object, and another frame of an image to be detected that includes the found second target object is obtained by using the second initial image. Therefore, performing target detection on the second initial image corresponding to the selected first initial image reduces the occupation of system resources to a certain extent.

其中，從若干幀第一初始圖像中，選擇第一目標對象符合活體檢測要求的第一初始圖像，作為最終第一初始圖像，包括：對於每幀第一初始圖像：基於第一目標對象的至少一個品質因數，得到第一目標對象的品質分數；其中，第一目標對象的品質因數包括以下至少一種：第一目標對象的置信度、角度、大小、模糊度以及第一目標對象所在的第一初始圖像的模糊度；選擇第一目標對象的品質分數大於預設分數閾值的第一初始圖像，作為最終第一初始圖像；利用最終第一初始圖像，得到包含查找出的第一目標對象的一幀待檢測圖像，包括：將最終第一初始圖像中的第一目標對象所在區域按照預設比例進行第一外擴，並提取第一外擴之後的區域作為一幀待檢測圖像；利用第二初始圖像，得到包含查找出的第二目標對象的另一幀待檢測圖像，包括：將第二初始圖像中的第二目標對象所在區域按照預設比例進行第二外擴，並提取第二外擴之後的區域作為另一幀待檢測圖像。因此，通過獲取品質分數滿足條件的第一初始圖像以及第一初始圖像對應的第二初始圖像，減輕了外界因素對活體檢測的影響，從而使得得到的待檢測圖像的活體檢測結果越準確；其中，通過對第一初始圖像中的第一目標對象以及對應第二初始圖像中的第二目標對象進行外擴提取，從而減輕了其他目標對象對活體檢測結果的影響，提高了活體檢測的精度。Wherein, from several frames of first initial images, selecting the first initial image of the first target object that meets the requirements of living body detection as the final first initial image, including: for each frame of the first initial image: based on the first At least one quality factor of the target object to obtain the quality score of the first target object; wherein, the quality factor of the first target object includes at least one of the following: the confidence, angle, size, ambiguity of the first target object and the first target object The ambiguity of the first initial image where it is located; the first initial image whose quality score of the first target object is greater than the preset score threshold is selected as the final first initial image; the final first initial image is used to obtain the A frame of the image to be detected of the first target object obtained, including: first expanding the area where the first target object is located in the final first initial image according to a preset ratio, and extracting the area after the first expanding As a frame of the image to be detected; using the second initial image to obtain another frame of the image to be detected containing the found second target object, including: placing the area where the second target object in the second initial image is located according to A second expansion is performed at a preset ratio, and an area after the second expansion is extracted as another frame of the image to be detected. Therefore, by acquiring the first initial image whose quality score satisfies the condition and the second initial image corresponding to the first initial image, the influence of external factors on the living body detection is alleviated, so that the obtained living body detection result of the image to be detected is obtained. The more accurate; wherein, by performing the expansion extraction on the first target object in the first initial image and the second target object in the corresponding second initial image, the influence of other target objects on the living body detection result is reduced, and the accuracy of live detection.

其中，對至少兩幀待檢測圖像進行活體檢測，得到關於目標對象的活體檢測結果，包括：對於每幀待檢測圖像，利用待檢測圖像對應的拍攝模式匹配的活體檢測模型對待檢測圖像進行活體檢測。因此，通過將拍攝模式與活體檢測模型對應，使得能夠更有針對性的對待檢測圖像進行活體檢測，使得活體檢測的準確度更高。Wherein, performing in vivo detection on at least two frames of the images to be detected to obtain the in vivo detection results of the target object, including: for each frame of the to-be-detected images, using the to-be-detected image to be detected by the in vivo detection model matched with the shooting mode corresponding to the to-be-detected images Like a liveness test. Therefore, by corresponding the shooting mode to the living body detection model, it is possible to perform the living body detection on the image to be detected in a more targeted manner, so that the accuracy of the living body detection is higher.

其中，活體檢測模型是樣本圖像訓練得到的；其中，活體檢測模型的樣本圖像是利用活體檢測模型匹配的拍攝模式拍攝得到的；樣本圖像包括對活體目標拍攝得到的活體樣本圖像和對假體目標拍攝得到的假體樣本圖像，假體目標包括二維靜態圖像、二維動態圖像、三維模具中的至少一種。因此，利用多樣的樣本對活體檢測模型進行訓練使得活體檢測模型的適用性更強，檢測結果的準確度更高。Among them, the living body detection model is obtained by training the sample images; wherein, the sample images of the living body detection model are obtained by using the shooting mode matched by the living body detection model; the sample images include the living body sample images obtained by shooting the living body target and The prosthesis sample image obtained by photographing the prosthesis target includes at least one of a two-dimensional static image, a two-dimensional dynamic image, and a three-dimensional mold. Therefore, using a variety of samples to train the in vivo detection model makes the in vivo detection model more applicable and the detection result more accurate.

其中，在對至少兩幀待檢測圖像進行活體檢測，得到關於目標對象的活體檢測結果之後，方法包括以下至少一步：在活體檢測結果為目標對象屬於活體的情況下，對其中一幀待檢測圖像進行目標識別，得到目標對象的識別結果；在活體檢測結果為目標對象不屬於活體的情況下，發送關於活體檢測結果的第一通知。因此，只有在活體檢測結果為活體的情況下進行目標識別，減輕了後續目標識別的計算量，當目標對象不屬於活體時，將檢測結果以第一通知的方式發送出去以使得後續能夠對檢測結果進行記錄。The method includes at least one of the following steps after performing in vivo detection on at least two frames of the images to be detected and obtaining the in vivo detection result about the target object: if the in vivo detection result is that the target object belongs to a living body, perform detection on one of the frames to be detected. The image is subjected to target recognition to obtain the recognition result of the target object; in the case that the target object does not belong to the living body as a result of the living body detection, a first notification about the living body detection result is sent. Therefore, target recognition is only performed when the living body detection result is a living body, which reduces the calculation amount of subsequent target recognition. When the target object does not belong to a living body, the detection result is sent as a first notification so that subsequent detection can be performed. The results are recorded.

其中，對其中一幀待檢測圖像進行目標識別，得到目標對象的識別結果，包括：對其中一幀待檢測圖像進行特徵提取，得到關於目標對象的目標特徵；獲取目標特徵分別與至少一個預存特徵的相似度；基於相似度，確定目標對象的識別結果。因此，通過相似度的比對來確定識別結果，使得目標識別結果更有依據且準確。Wherein, performing target recognition on one of the frames to be detected to obtain a recognition result of the target object includes: performing feature extraction on one of the frames to be detected to obtain target features about the target object; obtaining the target features respectively associated with at least one The similarity of the pre-stored features; based on the similarity, the recognition result of the target object is determined. Therefore, the recognition result is determined by comparing the similarity, so that the target recognition result is more basis and accurate.

其中，在得到目標對象的識別結果之後，方法還包括：在識別結果為目標對象被識別成功的情況下，執行與目標對象的身份匹配的聯動控制；在識別結果為目標對象未被識別成功的情況下，發送關於識別結果的第二通知。因此，通過將識別結果執行聯動控制使得聯動過程更加方便。Wherein, after the recognition result of the target object is obtained, the method further includes: if the recognition result is that the target object is successfully recognized, executing linkage control matching the identity of the target object; if the recognition result is that the target object has not been successfully recognized In this case, a second notification about the recognition result is sent. Therefore, the linkage process is more convenient by performing linkage control on the recognition result.

其中，執行與目標對象的身份匹配的聯動控制，包括：在目標對象的身份屬於第一類身份，則控制關聯的門體打開，和/或，將目標對象的身份發送給關聯的第一通信設備，以使第一通信設備基於目標對象的身份進行與第一類身份相關的業務；在目標對象的身份屬於第二類身份，則控制外接設備發出警報，和/或，將目標對象的身份發送給第一通信設備，以使第一通信設備基於目標對象的身份進行與第二類身份相關的業務；發送關於活體檢測結果的第一通知，包括：將待檢測圖像和活體檢測結果中的至少一者進行第一編碼，並將第一編碼的結果打包至第一通知以發送給第二通信設備；發送關於識別結果的第二通知，包括：將待檢測圖像和識別結果中的至少一者進行第二編碼，並將第二編碼的結果打包至第二通知以發送給第三通信設備。因此，通過目標識別結果聯動開門或聯動報警或其他相關業務，在一定程度上保障了通行的安全以及起到視頻監控的作用。Wherein, performing linkage control matching the identity of the target object includes: when the identity of the target object belongs to the first type of identity, controlling the opening of the associated door, and/or sending the identity of the target object to the associated first communication device, so that the first communication device can perform services related to the first type of identity based on the identity of the target object; when the identity of the target object belongs to the second type of identity, the external device is controlled to issue an alarm, and/or, the identity of the target object is Sending to the first communication device, so that the first communication device can perform services related to the second type of identity based on the identity of the target object; sending a first notification about the living body detection result, including: adding the image to be detected and the living body detection result in the At least one of the first encoding is performed, and the result of the first encoding is packaged into a first notification to be sent to the second communication device; sending a second notification about the recognition result includes: combining the image to be detected and the recognition result in the At least one performs the second encoding, and packages the result of the second encoding into the second notification for sending to the third communication device. Therefore, through the target recognition results, the linkage of door opening or linkage alarm or other related services can ensure the safety of traffic and play the role of video surveillance to a certain extent.

本發明實施例第二方面提供了一種拍攝處理裝置，包括：亮度獲取模組，配置為獲取拍攝設備所處環境的當前亮度資訊，其中，拍攝設備包括至少兩個圖像採集組件；模式切換模組，配置為將至少一個圖像採集組件調整為與當前亮度資訊匹配的拍攝模式，其中，不同拍攝模式是採用不同的光形成圖像。A second aspect of an embodiment of the present invention provides a photographing processing device, including: a brightness acquisition module configured to acquire current brightness information of an environment where a photographing device is located, wherein the photographing device includes at least two image acquisition components; a mode switching module The group is configured to adjust at least one image capturing component to a shooting mode matching the current brightness information, wherein different shooting modes use different light to form images.

本發明實施例第三方面提供了一種電子設備，包括記憶體和處理器，處理器配置為執行記憶體中儲存的程式指令，以實現上述拍攝處理方法。A third aspect of the embodiments of the present invention provides an electronic device, including a memory and a processor, where the processor is configured to execute program instructions stored in the memory, so as to implement the above-mentioned shooting processing method.

本發明實施例第四方面提供了一種電腦可讀儲存介質，其上儲存有程式指令，程式指令被處理器執行時實現上述拍攝處理方法。A fourth aspect of the embodiments of the present invention provides a computer-readable storage medium, on which program instructions are stored, and when the program instructions are executed by a processor, the above-mentioned shooting processing method is implemented.

本發明實施例第五方面提供了一種電腦程式，包括電腦可讀代碼，當所述電腦可讀代碼在電子設備中運行時，所述電子設備中的處理器執行配置為實現所述拍攝處理方法。A fifth aspect of the embodiments of the present invention provides a computer program, including computer-readable code, when the computer-readable code is executed in an electronic device, the execution of a processor in the electronic device is configured to implement the shooting processing method .

在本發明實施例中，通過使用至少兩個圖像採集組件拍攝得到圖像，相對於單攝影頭採集到的圖像，能夠提高圖像品質。而且，通過拍攝設備所處環境的當前亮度資訊來調整圖像採集組件的拍攝模式，使得能夠在不同的光照環境下，選擇合適的拍攝模式來採集得到圖像，以使採集到的圖像的品質更好。In the embodiment of the present invention, by using at least two image capture components to capture images, the image quality can be improved compared to images captured by a single camera. Moreover, the shooting mode of the image acquisition component is adjusted according to the current brightness information of the environment where the shooting device is located, so that an appropriate shooting mode can be selected to collect images under different lighting environments, so that the collected images can be Better quality.

應當理解的是，以上的一般描述和後文的細節描述僅是示例性和解釋性的，而非限制本發明。It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention.

下面結合說明書附圖，對本發明實施例的方案進行詳細說明。The solutions of the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

以下描述中，為了說明而不是為了限定，提出了諸如特定系統結構、介面、技術之類的具體細節，以便透徹理解本發明實施例。In the following description, for the purpose of illustration rather than limitation, specific details, such as specific system structures, interfaces, and technologies, are set forth in order to provide a thorough understanding of the embodiments of the present invention.

本文中術語“和/或”，僅僅是一種描述關聯對象的關聯關係，表示可以存在三種關係，例如，A和/或B，可以表示：單獨存在A，同時存在A和B，單獨存在B這三種情況。另外，本文中字元“/”，一般表示前後關聯對象是一種“或”的關係。此外，本文中的“多”表示兩個或者多於兩個。另外，本文中術語“至少一種”表示多種中的任意一種或多種中的至少兩種的任意組合，例如，包括A、B、C中的至少一種，可以表示包括從A、B和C構成的集合中選擇的任意一個或多個元素。The term "and/or" in this article is only an association relationship to describe associated objects, indicating that there can be three kinds of relationships, for example, A and/or B, which can mean that A exists alone, A and B exist at the same time, and B exists alone. three situations. In addition, the character "/" in this text generally indicates that the related objects are an "or" relationship. Also, "multiple" herein means two or more than two. In addition, the term "at least one" herein refers to any combination of any one of a plurality or at least two of a plurality, for example, including at least one of A, B, and C, and may mean including those composed of A, B, and C. Any one or more elements selected in the collection.

請參閱圖1，圖1是本發明實施例提供的拍攝處理方法的流程示意圖。可以包括如下步驟。Please refer to FIG. 1. FIG. 1 is a schematic flowchart of a photographing processing method provided by an embodiment of the present invention. The following steps may be included.

步驟S11：獲取拍攝設備所處環境的當前亮度資訊，其中，拍攝設備包括至少兩個圖像採集組件。Step S11: Acquire current brightness information of the environment where the photographing device is located, wherein the photographing device includes at least two image acquisition components.

一些公開實施例中，圖像採集組件包括鏡頭以及圖像感測器。In some disclosed embodiments, the image capture assembly includes a lens and an image sensor.

在一些實施方式中，獲取拍攝設備所處環境的當前亮度資訊的方式可以包括但不局限於以下兩種，方式一：獲取當前圖像採集組件拍攝的圖像，提取拍攝的圖像的亮度資訊，作為當前拍攝設備所處環境的亮度資訊；方式二，可以在拍攝設備上設置光敏組件，利用光敏組件獲取到的亮度資訊作為拍攝設備所處環境的當前亮度資訊。本發明實施例選擇方式一獲取拍攝設備所述環境的當前亮度資訊，拍攝得到的圖像的亮度直接影響到圖像的品質，因此，通過從高拍攝得到的圖像中獲取亮度資訊作為環境的亮度資訊更能夠體現當前環境亮度對拍攝圖像的影響程度。當然在其他公開實施例中也可單獨採用方式二獲取拍攝設備所處環境的當前亮度資訊，也可將方式一與方式二結合得到拍攝設備所處環境的當前亮度資訊，此時，可以設置方式一以及方式二獲取到的環境亮度資訊的權重，例如，通過預先設置方式一獲取到的環境亮度資訊的權重值為三分之二，設置方式二獲得的環境亮度資訊的權重值為三分之一，然後將方式一以及方式二獲取到的環境亮度資訊乘以各自的權重值然後相加得到最終的當前環境亮度資訊。因此，在實施過程中獲取拍攝設備所處環境的當前亮度資訊的方式此處不做限定。In some embodiments, the methods of acquiring the current brightness information of the environment where the photographing device is located may include, but are not limited to, the following two methods. Method 1: acquiring the image captured by the current image capture component, and extracting the brightness information of the captured image. , as the current brightness information of the environment where the photographing device is located; in the second method, a photosensitive component can be set on the photographing device, and the brightness information obtained by the photosensitive component can be used as the current brightness information of the environment where the photographing device is located. The first option of the embodiment of the present invention is to obtain the current brightness information of the environment described by the photographing device. The brightness of the captured image directly affects the quality of the image. The brightness information can better reflect the influence of the current ambient brightness on the captured image. Of course, in other disclosed embodiments, the second method can be used alone to obtain the current brightness information of the environment where the photographing device is located, or the first method and the second method can be combined to obtain the current brightness information of the environment where the photographing device is located. The weight of the ambient brightness information obtained in the first and second methods, for example, the weight value of the ambient brightness information obtained by the preset method 1 is two-thirds, and the weight value of the ambient brightness information obtained by setting the second method is one-third. First, the ambient brightness information obtained in the first and second methods is multiplied by the respective weight values and then added together to obtain the final current ambient brightness information. Therefore, the manner of acquiring the current brightness information of the environment where the photographing device is located in the implementation process is not limited here.

步驟S12：將至少一個圖像採集組件調整為與當前亮度資訊匹配的拍攝模式；其中，不同拍攝模式是採用不同的光形成圖像。Step S12: Adjust at least one image capturing component to a shooting mode matching the current brightness information; wherein, different shooting modes use different lights to form images.

一些公開實施例中，一個圖像採集組件可以設置多種拍攝模式，例如兩種，不同的拍攝模式採用不同的光形成圖像，其中，光可以是自然光，可以是紅外光等等，此處不做限定。將至少一個圖像採集組件調整為與當前亮度資訊匹配的拍攝模式的方式，其中，這裡的當前亮度資訊可以是一個亮度區間，可以是一個確切的亮度值。在一些實施方式中，這裡的亮度區間可以僅具有最小亮度值，最大亮度值不限的半包圍區間，也可以是僅具有最大亮度值，最小亮度值不限的半包圍區間。本發明實施例中當前亮度資訊指的是一個亮度區間。通過當前亮度資訊調整拍攝模式的方式可以是當環境亮度資訊處於第一環境亮度區間時，則控制其中一個或兩個圖像採集組件調整為與第一環境亮度區間匹配的拍攝模式，而當環境亮度資訊處於第二環境亮度區間時，則將其中一個或兩個圖像採集組件調整為與第二環境亮度區間匹配的拍攝模式。In some disclosed embodiments, one image capture component can be set with multiple shooting modes, such as two, different shooting modes use different light to form images, wherein the light can be natural light, infrared light, etc. Do limit. A method of adjusting at least one image acquisition component to a shooting mode matching the current brightness information, wherein the current brightness information here may be a brightness interval, or an exact brightness value. In some embodiments, the brightness interval here may be a semi-enclosed interval with only a minimum brightness value and an unlimited maximum brightness value, or a semi-enclosed interval with only a maximum brightness value and an unlimited minimum brightness value. In the embodiment of the present invention, the current luminance information refers to a luminance interval. The method of adjusting the shooting mode through the current brightness information may be that when the ambient brightness information is in the first ambient brightness interval, one or two of the image capture components are controlled to be adjusted to the shooting mode matching the first ambient brightness interval, and when the ambient brightness information is in the first ambient brightness interval When the brightness information is in the second ambient brightness interval, one or two of the image capturing components are adjusted to a shooting mode matching the second ambient brightness interval.

上述方案，通過使用至少兩個圖像採集組件拍攝得到圖像，相對於單攝影頭採集到的圖像，能夠提高圖像品質。而且，通過拍攝設備所處環境的當前亮度資訊來調整圖像採集組件的拍攝模式，使得能夠在不同的光照環境下，選擇合適的拍攝模式來採集得到圖像，以使採集到的圖像的品質更好。In the above solution, by using at least two image capturing components to capture images, the image quality can be improved compared to images captured by a single camera. Moreover, the shooting mode of the image acquisition component is adjusted according to the current brightness information of the environment where the shooting device is located, so that an appropriate shooting mode can be selected to collect images under different lighting environments, so that the collected images can be Better quality.

一些公開實施例中，至少兩個圖像採集組件可以是包括兩個圖像採集組件，也可以是包括三個圖像採集組件等等。兩個圖像採集組件的拍攝設備可以是雙目攝影機，三個圖像採集組件的拍攝設備可以是三目攝影機。本發明實施例以拍攝設備包括兩個圖像採集組件為例，當然，在其他實施例中，拍攝設備可以包括三個或三個以上的圖像採集組件。In some disclosed embodiments, the at least two image capturing assemblies may include two image capturing assemblies, or may include three image capturing assemblies, or the like. The photographing devices of the two image capturing assemblies may be binocular cameras, and the photographing devices of the three image capturing assemblies may be trinocular cameras. The embodiment of the present invention takes the photographing device including two image capturing components as an example. Of course, in other embodiments, the photographing device may include three or more image capturing components.

參見圖2，圖2是本發明實施例提供的拍攝處理方法中具有兩圖像採集組件的拍攝設備的結構示意圖。如圖2所示，拍攝設備20具有第一圖像採集組件21和第二圖像採集組件22，其中，第一圖像採集組件21和第二圖像採集組件22可並列設置在拍攝設備20上。圖2中，拍攝設備20的第一圖像採集組件21和第二圖像採集組件22的基線距離大於等於60毫米（mm）。當然，在其他實施例中，兩個圖像採集組件還可處於同一豎直方向上。因此，第一圖像採集組件21和第二圖像採集組件22在拍攝設備20上如何排列，本發明實施例不做限定。Referring to FIG. 2, FIG. 2 is a schematic structural diagram of a photographing device having two image acquisition components in a photographing processing method provided by an embodiment of the present invention. As shown in FIG. 2 , the photographing device 20 has a first image capturing assembly 21 and a second image capturing assembly 22 , wherein the first image capturing assembly 21 and the second image capturing assembly 22 can be arranged side by side on the photographing device 20 superior. In FIG. 2 , the distance between the baselines of the first image capturing assembly 21 and the second image capturing assembly 22 of the photographing device 20 is greater than or equal to 60 millimeters (mm). Of course, in other embodiments, the two image capturing assemblies may also be located in the same vertical direction. Therefore, how the first image capturing assembly 21 and the second image capturing assembly 22 are arranged on the photographing device 20 is not limited in this embodiment of the present invention.

同一拍攝設備上的圖像採集組件均是幀同步的，其中，可以對圖像採集組件進行分類，例如，將其中一個圖像採集組件作為主圖像採集組件，另外所有的圖像採集組件作為副圖像採集組件。主圖像採集組件與本發明實施例中的第一圖像採集組件相同，所有的副圖像採集組件與第二圖像採集組件相同。至少兩個圖像採集組件包括第一圖像採集組件和第二圖像採集組件。其中，第一圖像採集組件和第二圖像採集組件是幀同步的。圖像採集組件的拍攝模式包括第一拍攝模式和第二拍攝模式。在一些實施方式中，第一拍攝模式可以是彩色拍攝模式，第二拍攝模式可以是紅外拍攝模式。其中，彩色拍攝模式則是利用可見光進行成像，而紅外拍攝模式則是利用紅外光進行成像。The image acquisition components on the same shooting device are all frame-synchronized, wherein the image acquisition components can be classified, for example, one of the image acquisition components is used as the main image acquisition component, and all other image acquisition components are Secondary image acquisition component. The main image acquisition component is the same as the first image acquisition component in the embodiment of the present invention, and all the secondary image acquisition components are the same as the second image acquisition component. The at least two image capture assemblies include a first image capture assembly and a second image capture assembly. Wherein, the first image acquisition component and the second image acquisition component are frame-synchronized. The shooting modes of the image capturing assembly include a first shooting mode and a second shooting mode. In some embodiments, the first photographing mode may be a color photographing mode, and the second photographing mode may be an infrared photographing mode. Among them, the color shooting mode uses visible light for imaging, and the infrared shooting mode uses infrared light for imaging.

其中，在當前亮度資訊大於亮度閾值的情況下，將至少一個圖像採集組件調整為第一拍攝模式。其中，這裡將圖像採集組件調整為第一拍攝模式的實施方式可以是調整圖像採集組件中圖像感測器的模式為第一拍攝模式。也就是經過圖像感測器生成的是第一拍攝模式對應的圖像，當第一拍攝模式為彩色模式時，圖像感測器生成的圖像為彩色圖像。Wherein, when the current brightness information is greater than the brightness threshold, at least one image acquisition component is adjusted to the first shooting mode. Wherein, the implementation manner of adjusting the image capture component to the first shooting mode here may be to adjust the mode of the image sensor in the image capture component to the first shooting mode. That is, the image corresponding to the first shooting mode is generated by the image sensor. When the first shooting mode is the color mode, the image generated by the image sensor is a color image.

在一些實施方式中，拍攝設備中的每個圖像採集組件的拍攝模式相同。兩個圖像採集組件的拍攝模式相同，說明當亮度資訊大於亮度閾值時，將兩個圖形採集組件皆調整為第一拍攝模式，當然，如果在調整之前兩個圖像採集組件就處於第一拍攝模式下，則繼續保持第一拍攝模式即可；當亮度資訊小於或等於亮度閾值時，則將兩個圖像採集組件皆調整為第二拍攝模式，當然，如果在調整之前兩個圖像採集組件就處於第二拍攝模式下，則繼續保持第二拍攝模式即可。In some embodiments, the capture mode of each image capture assembly in the capture device is the same. The shooting modes of the two image acquisition components are the same, which means that when the brightness information is greater than the brightness threshold, the two image acquisition components are adjusted to the first shooting mode. Of course, if the two image acquisition components are in the first shooting mode before the adjustment In the shooting mode, the first shooting mode can be maintained; when the brightness information is less than or equal to the brightness threshold, the two image acquisition components are adjusted to the second shooting mode. As long as the acquisition component is in the second shooting mode, the second shooting mode can be maintained.

上述方案，通過在亮度資訊大於閾值時，選擇彩色拍攝模式，當亮度資訊小於閾值時，選擇紅外拍攝模式，使得當拍攝設備在白天亮度較高紅外線過強時，減少紅外成像會受到強紅外線的干擾。而在夜晚亮度較暗時，選擇紅外拍攝模式，減輕了對可見光的依賴，從而提高了成像品質。In the above solution, when the brightness information is greater than the threshold value, the color shooting mode is selected, and when the brightness information is less than the threshold value, the infrared shooting mode is selected, so that when the brightness of the shooting device is high during the day and the infrared rays are too strong, the infrared imaging will be less affected by the strong infrared rays. interference. When the brightness is dark at night, the infrared shooting mode is selected to reduce the dependence on visible light, thereby improving the imaging quality.

一些公開實施例中，當圖像採集組件調整為第一拍攝模式時，也就是調整為彩色模式時，關閉拍攝設備的紅外補光燈，因為當圖像採集組件為彩色模式，利用可見光進行形成圖像，因此，無需打開紅外補光燈進行補光。同樣，當圖像採集組件調整為第二拍攝模式時，即調整為紅外拍攝模式時，則打開拍攝設備的紅外補光燈，打開了拍攝設備的紅外補光燈相當於開了一盞紅外線燈，照亮了拍攝設備的拍攝區域，然後拍攝設備通過接收紅外線的反光形成圖像。In some disclosed embodiments, when the image capture assembly is adjusted to the first shooting mode, that is, to the color mode, the infrared fill light of the shooting device is turned off, because when the image capture assembly is in the color mode, visible light is used to form the image. image, therefore, there is no need to turn on the infrared fill light for fill light. Similarly, when the image capture component is adjusted to the second shooting mode, that is, to the infrared shooting mode, the infrared fill light of the shooting device is turned on. Turning on the infrared fill light of the shooting device is equivalent to turning on an infrared light. , illuminates the shooting area of the shooting device, and then the shooting device forms an image by receiving the reflected light of infrared rays.

其中，當至少圖像採集組件調整為第一拍攝模式時，將拍攝設備的紅外截止濾光片移動至需調整為第一拍攝模式的圖像採集組件的入光通道中。入光通道指的是鏡頭與圖像感測器之間的通道，即光線從鏡頭到圖像感測器所經過的通道。其中，若只有一個圖像採集組件需要調整為第一拍攝模式，那麼紅外截止濾光片則移動到該圖像採集組件的入光通道中，能夠過濾外界的紅外光，減輕紅外光對拍攝的圖像的影響；若兩個圖像採集組件均需要調整為第一拍攝模式，那麼兩個圖像採集組件的入光通道均有紅外截止濾光片。而當至少一個圖像採集組件調整為第二拍攝模式時，則將拍攝設備的紅外截止濾光片移動至需調整為第二拍攝模式的圖像採集組件的入光通道以外。其意思是，當其中一個圖像採集組件調整為第二拍攝模式時，則將拍攝設備的紅外截止濾光片移動至該圖像採集組件的入光通道以外，不對紅外光進行過濾，使得圖像採集組件可以利用紅外光進行成像。Wherein, when at least the image acquisition component is adjusted to the first shooting mode, the infrared cut filter of the shooting device is moved to the light incident channel of the image acquisition component that needs to be adjusted to the first shooting mode. The light entrance channel refers to the channel between the lens and the image sensor, that is, the channel through which light passes from the lens to the image sensor. Among them, if only one image acquisition component needs to be adjusted to the first shooting mode, the infrared cut-off filter is moved to the light incident channel of the image acquisition component, which can filter the infrared light from the outside world and reduce the effect of infrared light on shooting. The influence of the image; if both image acquisition components need to be adjusted to the first shooting mode, the light incident channels of the two image acquisition components have infrared cut-off filters. When at least one image capture component is adjusted to the second shooting mode, the infrared cut filter of the shooting device is moved outside the light incident channel of the image capture component that needs to be adjusted to the second shooting mode. It means that when one of the image acquisition components is adjusted to the second shooting mode, the infrared cut-off filter of the shooting device is moved out of the light incident channel of the image acquisition component, and the infrared light is not filtered, so that the image is not filtered. Image acquisition components can use infrared light for imaging.

上述方案，在圖像採集組件為彩色拍攝模式時，說明環境亮度較高，紅外光比較強，因此，無需打開紅外補光燈，且將紅外截止濾光片放置在圖像採集組件的入光通道內，對紅外光進行過濾，減輕了紅外光對拍攝到的圖像的影響。而在圖像採集組件為紅外拍攝模式時，說明環境亮度較暗，才利用打開紅外補光燈，使得拍攝到的圖像更清楚即品質更好。In the above solution, when the image acquisition component is in color shooting mode, it means that the ambient brightness is high and the infrared light is relatively strong. Therefore, it is not necessary to turn on the infrared fill light, and the infrared cut-off filter is placed on the incoming light of the image acquisition component. In the channel, the infrared light is filtered to reduce the influence of the infrared light on the captured image. When the image acquisition component is in the infrared shooting mode, it means that the ambient brightness is dark, and the infrared fill light is turned on, so that the captured image is clearer, that is, the quality is better.

一些公開實施例中，拍攝處理方法包括，獲取至少兩幀初始圖像，其中，至少兩幀初始圖像為兩個圖像採集組件對所處環境中的目標對象分別拍攝得到的。本發明實施例中，將第一圖像採集組件拍攝得到的初始圖像稱為第一初始圖像，第二圖像採集組件拍攝得到的初始圖像稱為第二初始圖像。一些公開實施例中，將可以對第一初始圖像和第二初始圖像進行編碼以便後續儲存，或者發送到其他關聯設備。In some disclosed embodiments, the photographing processing method includes acquiring at least two frames of initial images, wherein the at least two frames of initial images are obtained by separately photographing a target object in an environment by two image acquisition components. In the embodiment of the present invention, the initial image captured by the first image capturing component is referred to as the first initial image, and the initial image captured by the second image capturing module is referred to as the second initial image. In some disclosed embodiments, it will be possible to encode the first initial image and the second initial image for subsequent storage or transmission to other associated devices.

一些公開實施例中，利用獲取到的第一初始圖像和/或第二初始圖像獲取攝影設備所處環境的當前亮度資訊，然後將其中一個或兩個圖像採集組件調整為與當前亮度資訊匹配的拍攝模式，以使得圖像採集組件在後續利用調整後的拍攝模式獲取圖像。其中，第一圖像採集組件和第二圖像採集組件的拍攝模式包括第一拍攝模式和第二拍攝模式，其中，可根據環境的亮度資訊來動態調整第一圖像採集組件和第二圖像採集組件的拍攝模式。在實施中調整拍攝模式的方式如上所述，此處不再贅述。當然，在其他實施例中，也可在獲取初始圖像之前，先行調整拍攝設備中第一圖像採集組件和/或第二圖像採集組件的拍攝模式，例如利用在獲取初始圖像之前，利用光敏組件採集到的亮度資訊，來調整拍攝設備的第一圖像採集組件和/或第二圖像採集組件的拍攝模式。In some disclosed embodiments, the obtained first initial image and/or the second initial image is used to obtain the current brightness information of the environment where the photographing device is located, and then one or both of the image capture components are adjusted to match the current brightness. information matching the shooting mode, so that the image acquisition component uses the adjusted shooting mode to acquire images subsequently. Wherein, the shooting modes of the first image acquisition component and the second image acquisition component include a first shooting mode and a second shooting mode, wherein the first image acquisition component and the second image acquisition component can be dynamically adjusted according to the brightness information of the environment. Like the capture mode of the capture component. The manner of adjusting the shooting mode in the implementation is as described above, and will not be repeated here. Of course, in other embodiments, before acquiring the initial image, the shooting mode of the first image acquisition component and/or the second image acquisition component in the photographing device may be adjusted first, for example, by using, before acquiring the initial image, The brightness information collected by the photosensitive component is used to adjust the shooting mode of the first image acquisition component and/or the second image acquisition component of the photographing device.

基於每幀初始圖像，對應得到包含目標對象的待檢測圖像。其中，待檢測圖像可用於活體檢測。在一些實施方式中，從第一圖像採集組件採集到的若干幀第一初始圖像中，選出最終第一初始圖像。其中，第一圖像採集組件對目標對象拍攝若干幀初始圖像，可以從這些若干幀初始圖像中選擇出合適的一幀或多幀初始圖像作為最終第一初始圖像。Based on the initial image of each frame, the to-be-detected image containing the target object is correspondingly obtained. Among them, the image to be detected can be used for living body detection. In some implementations, the final first initial image is selected from several frames of the first initial images acquired by the first image acquisition component. Wherein, the first image acquisition component captures several frames of initial images of the target object, and can select a suitable frame or frames of initial images from these several frames of initial images as the final first initial image.

一些公開實施例中，從第一圖像採集組件對目標對象拍攝得到的若干幀第一初始圖像中，選出最終第一初始圖像的方式可以是：先對每幀第一初始圖像進行目標檢測和跟蹤，得到每幀第一初始圖像所包含的第一目標對象。在一些實施方式中，對每幀第一初始圖像進行目標檢測和跟蹤，得到目標檢測結果，其中，目標檢測結果中包括第一目標對象以及第一目標對象的關鍵點資訊。其中，第一目標對象即是需要進行活體檢測或目標識別的對象，例如人臉。其中，這裡的關鍵點資訊指的是第一目標對象中具有可識別性的點的資訊，例如當第一目標對象為人臉時，關鍵點的資訊可以是具有可識別性的眉毛、眼睛、鼻子或嘴巴等關鍵點的資訊。在其他公開實施例中，在對第一初始圖像進行目標檢測和跟蹤之前，可以先對獲取到的第一初始圖像和第二初始圖像進行預處理，其中，預處理的方式包括調整第一初始圖像和第二初始圖像的尺寸和/或色彩，使得後續能夠對統一樣式的第一初始圖像和第二初始圖像進行目標檢測和跟蹤，使得目標檢測和跟蹤的魯棒性更強，目標檢測結果更準確。In some disclosed embodiments, the manner of selecting the final first initial image from several frames of the first initial images captured by the first image acquisition component of the target object may be: Target detection and tracking, to obtain the first target object included in the first initial image of each frame. In some embodiments, target detection and tracking are performed on the first initial image of each frame to obtain a target detection result, wherein the target detection result includes the first target object and key point information of the first target object. The first target object is an object that needs to be subjected to live detection or target recognition, such as a human face. Wherein, the key point information here refers to the information of the identifiable points in the first target object. For example, when the first target object is a human face, the information of the key points may be identifiable eyebrows, eyes, Information on key points such as the nose or mouth. In other disclosed embodiments, before performing target detection and tracking on the first initial image, preprocessing may be performed on the acquired first initial image and the second initial image, wherein the preprocessing method includes adjusting The size and/or color of the first initial image and the second initial image, so that the target detection and tracking can be performed on the first initial image and the second initial image of the uniform style subsequently, so that the target detection and tracking are robust. The performance is stronger, and the target detection result is more accurate.

對每幀述第一初始圖像進行目標檢測和跟蹤，得到每幀述第一初始圖像所包含的第一目標對象之後，從若干幀第一初始圖像中，選擇第一目標對象符合活體檢測要求的第一初始圖像，作為最終第一初始圖像。在實施中選擇方式可以是對於每幀第一初始圖像：基於第一目標對象的至少一個品質因數，得到第一目標對象的品質分數，其中第一目標對象的品質因數包括以下至少一種：第一目標對象的置信度、角度、大小、模糊度以及第一目標對象所在的第一初始圖像的模糊度。其中這裡的第一目標對象的置信度指的是檢測到的第一目標對象是真的第一目標對象的概率，例如當第一目標對象為人臉時，對第一初始圖像進行檢測時，可能會將其他動物的臉認定為人臉的概率為0.6，此時，這裡的0.6就是其他動物人臉的置信度。角度指的是第一目標對象相對於第一圖像採集組件的角度，例如可分為XYZ方向的角度。例如，以拍攝設備中第一圖像採集組件作為原點，建立三維坐標系，例如，拍攝設備的第一圖像採集組件與地心的連線為X軸，拍攝設備中兩圖像採集組件與X軸垂直的連線作為Y軸，過原點且同時與X軸和Y軸垂直的連線作為Z軸，當然，這個三維坐標系僅為舉例，三維坐標系還可以其他物體作為原點建立，以此來計算拍攝設備與目標對象的角度關係。例如，第一目標對象正對第一圖像採集組件則沿XYZ方向上的角度皆為0°，而第一目標對象正側對第一圖像採集組件，則第一目標對象相對於第一圖像採集組件X方向上的角度為90°，沿Y方向上的角度為0°，沿Z方向上的角度也為0°。當然，在其他實施例中，也可不必建立三維坐標系，其他能用於表示拍攝對象和拍攝設備之間的角度的方式均可，例如，先判斷第一目標對象是否被其他物體遮擋，若沒有被遮擋，則提取第一目標對象的關鍵點，將從多個目標對象上提取到的關鍵點進行比較；若關鍵點的數量較多，則認為該目標對象與拍攝設備的角度越小，反之則認為該目標對象與拍攝設備的角度越大。大小指的是第一目標對象所占第一初始圖像的面積大小，一般地，第一目標對象遠離第一圖像採集組件，則面積越小，越靠近第一圖像採集組件則面積越大。面積越大則活體檢測結果越準確，若面積過小，則對活體檢測的影響越大，檢測結果可能不是很準確。模糊度則指的是第一目標對象或第一初始圖像的模糊程度。Perform target detection and tracking on each frame of the first initial image, and after obtaining the first target object included in each frame of the first initial image, from several frames of the first initial image, select the first target object that conforms to the living body The required first initial image is detected as the final first initial image. In the implementation, the selection method may be, for each frame of the first initial image: obtaining a quality score of the first target object based on at least one quality factor of the first target object, wherein the quality factor of the first target object includes at least one of the following: The confidence, angle, size, and ambiguity of a target object, and the ambiguity of the first initial image where the first target object is located. The confidence of the first target object here refers to the probability that the detected first target object is the real first target object. For example, when the first target object is a human face, when the first initial image is detected , the probability of identifying the faces of other animals as human faces is 0.6. At this time, 0.6 here is the confidence level of other animal faces. The angle refers to the angle of the first target object relative to the first image capturing component, and can be divided into angles in the XYZ directions, for example. For example, a three-dimensional coordinate system is established with the first image acquisition component in the photographing device as the origin. The line perpendicular to the X axis is used as the Y axis, and the line passing through the origin and perpendicular to both the X axis and the Y axis is used as the Z axis. Of course, this three-dimensional coordinate system is only an example, and other objects can also be used as the origin of the three-dimensional coordinate system. It is established to calculate the angular relationship between the shooting device and the target object. For example, if the first target object is facing the first image acquisition component, the angles along the XYZ directions are all 0°, and the first target object is facing the first image acquisition component on the side, the first target object is relative to the first The angle of the image acquisition component in the X direction is 90°, the angle along the Y direction is 0°, and the angle along the Z direction is also 0°. Of course, in other embodiments, it is not necessary to establish a three-dimensional coordinate system, and other methods can be used to represent the angle between the shooting object and the shooting device. For example, first determine whether the first target object is blocked by other objects, if If it is not blocked, the key points of the first target object are extracted, and the key points extracted from multiple target objects are compared; if the number of key points is large, it is considered that the angle between the target object and the shooting device is smaller, On the contrary, it is considered that the angle between the target object and the shooting device is larger. The size refers to the size of the area of the first initial image occupied by the first target object. Generally, the first target object is farther away from the first image acquisition component, the smaller the area, and the closer it is to the first image acquisition component, the larger the area. big. The larger the area is, the more accurate the living body detection result will be. If the area is too small, the greater the impact on the living body detection will be, and the detection result may not be very accurate. The blur degree refers to the blur degree of the first target object or the first initial image.

其中，通過多個品質因數來計算品質分數時，可以預設各個品質因數所占權重，例如以第一目標對象的置信度以及角度、大小來計算品質分數時，預設置信度的權重為0.4、角度的權重為0.3、大小的權重為0.3，當然，這只是舉例，利用哪些品質因數來計算品質分數本發明實施例不做限定，權重的設置可以考慮到實際的圖像檢測精度需求以及圖像檢測設備的處理能力、資源佔用情況等。例如，一些公開實施例中，若圖像檢測設備的處理能力較高、資源佔用較少，則可考量多個品質因數來計算品質分數，而如果圖像檢測設備的處理能力過低，則可適當採用幾個品質因數來計算品質分，例如根據計算各個品質因數的所需時間或記憶體空間佔用來選擇合適的品質因數。因此，採用多少品質因數或採用哪幾個品質因數，可靈活做出選擇。Wherein, when calculating the quality score by using multiple quality factors, the weight occupied by each quality factor can be preset. For example, when calculating the quality score based on the confidence, angle, and size of the first target object, the preset reliability weight is 0.4 , the weight of the angle is 0.3, and the weight of the size is 0.3. Of course, this is just an example, which quality factor is used to calculate the quality score is not limited in this embodiment of the present invention, and the setting of the weight can take into account the actual image detection accuracy requirements and graphs. Such as detecting the processing capacity of the device, resource occupancy, etc. For example, in some disclosed embodiments, if the processing capability of the image detection device is high and the resource occupation is low, multiple quality factors may be considered to calculate the quality score, and if the processing capability of the image detection device is too low, the quality score may be calculated Several quality factors are appropriately used to calculate the quality score, for example, an appropriate quality factor is selected according to the time required to calculate each quality factor or the memory space occupied. Therefore, the choice of how many quality factors to use or which quality factors to use can be flexibly made.

得到第一目標對象的品質分數之後，選擇第一目標對象的品質分數大於預設分數閾值的第一初始圖像，作為最終第一初始圖像。例如，預設分數閾值為0.7、0.8或0.9等。經檢測，如果發現若干幀中有多幀第一初始圖像皆滿足活體檢測要求，可以從滿足活體檢測要求的若干幀中選擇出品質分數最高的一幀作為最終第一初始圖像。當然，在其他實施例中，將滿足活體檢測要求的第一初始圖像全部或部分作為最終第一初始圖像均可，可按照需求選擇實際的處理方式。當然，在其他實施例中，也可以是確定一個較低的品質分數閾值，若第一圖像的品質分數低於品質分數閾值，就將其排除，保留品質分數大於該品質分數閾值的第一圖像。After the quality score of the first target object is obtained, the first initial image whose quality score of the first target object is greater than the preset score threshold is selected as the final first initial image. For example, the preset score threshold is 0.7, 0.8, or 0.9, etc. After detection, if it is found that multiple first initial images in several frames meet the requirements of living body detection, a frame with the highest quality score can be selected as the final first initial image from several frames that meet the requirements of living body detection. Of course, in other embodiments, all or part of the first initial image that meets the requirements of living body detection may be used as the final first initial image, and an actual processing method can be selected according to requirements. Of course, in other embodiments, a lower quality score threshold may also be determined. If the quality score of the first image is lower than the quality score threshold, it will be excluded, and the first image whose quality score is greater than the quality score threshold will be retained. image.

獲取第二圖像採集組件採集的與最終第一初始圖像對應幀的第二初始圖像。如上所述，本發明實施例中描述的第一圖像採集組件與第二圖像採集組件幀同步。也就是只要第一圖像採集組件與第二圖像採集組件能夠正常工作，那麼當第一圖像採集組件在第一時刻獲取到一幀圖像時，第二圖像採集組件也會在第一時刻同步獲取一幀圖像。Acquire a second initial image of a frame corresponding to the final first initial image captured by the second image capture component. As described above, the first image capture component described in the embodiments of the present invention is frame-synchronized with the second image capture component. That is, as long as the first image acquisition component and the second image acquisition component can work normally, when the first image acquisition component acquires a frame of image at the first moment, the second image acquisition component will also Acquire one frame of image synchronously at one time.

其中，若選擇出了一幀或多幀最終第一初始圖像，那麼會在第二初始圖像中選擇與這些最終第一初始圖像幀同步的第二初始圖像，作為最終第二初始圖像。例如，經過篩選發現若干幀第一初始圖像中就第一幀的第一初始圖像滿足活體檢測的要求，那麼獲取第二圖像採集組件採集的與該第一幀第一初始圖像對應幀的第二初始圖像，也就是第二圖像採集組件採集的第一幀第二初始圖像。Wherein, if one or more frames of the final first initial image are selected, a second initial image synchronized with these final first initial image frames will be selected in the second initial image as the final second initial image image. For example, after screening, it is found that the first initial image of the first frame meets the requirements of living body detection among several frames of the first initial image, then the first initial image of the first frame corresponding to the first frame of the first initial image collected by the second image acquisition component is obtained. The second initial image of the frame, that is, the second initial image of the first frame acquired by the second image acquisition component.

通過獲取品質分數滿足條件的第一初始圖像以及第一初始圖像對應的第二初始圖像，減輕了外界因素對活體檢測的影響，從而使得後續得到的待檢測圖像的活體檢測結果越準確。By acquiring the first initial image whose quality score satisfies the condition and the second initial image corresponding to the first initial image, the influence of external factors on the living body detection is alleviated, so that the living body detection result of the subsequently obtained image to be detected is more accurate. precise.

分別利用最終第一初始圖像和第二初始圖像，對應得到包含目標對象的兩幀待檢測圖像。這裡的兩幀待檢測圖像指的是最終第一初始圖像會對應得到一幀待檢測圖像，與最終第一初始圖像幀同步的最終第二初始圖像會對應得到另一幀待檢測圖像。當然，如果在選擇出的最終第一初始圖像與第二初始圖像只有一組時，利用幀同步的最終第一初始圖像和最終第二初始圖像得到一組待檢測圖像，若存在多組最終第一初始圖像和第二初始圖像，那麼就會對應得到多組包含目標對象的兩幀待檢測圖像。Using the final first initial image and the second initial image respectively, two frames of to-be-detected images containing the target object are correspondingly obtained. The two frames of images to be detected here refer to the fact that the final first initial image will correspondingly obtain one frame of the to-be-detected image, and the final second initial image synchronized with the final first initial image frame will correspondingly obtain another frame to be detected. Detect images. Of course, if there is only one set of the final first initial image and the second initial image selected, a set of images to be detected is obtained by using the final first initial image and the final second initial image of frame synchronization. If there are multiple sets of final first initial images and second initial images, then multiple sets of two frames of images to be detected containing the target object will be obtained correspondingly.

通過先在第一圖像採集組件採集到的若干張第一初始圖像，然後再利用選擇出的第一初始圖像以及對應的第二初始圖像獲得待檢測圖像，加快了選擇的速率，在一定程度上提高了獲取待檢測圖像的效率。而且相比於對兩個圖像採集組件的初始圖像分別進行選擇，本方案只需對其中一個圖像採集組件的初始圖像進行選擇，可以減少對處理資源的使用。By first acquiring several first initial images collected by the first image acquisition component, and then using the selected first initial image and the corresponding second initial image to obtain the image to be detected, the speed of selection is accelerated , to a certain extent, the efficiency of acquiring the image to be detected is improved. Moreover, compared with the selection of the initial images of the two image acquisition components, the solution only needs to select the initial images of one of the image acquisition components, which can reduce the use of processing resources.

在一些實施方式中對第二初始圖像進行目標檢測，得到第二初始圖像所包含的第二目標對象。其中，這裡的第二初始圖像指的是與選擇出的最終第一初始圖像對應的第二初始圖像，而不是第二圖像採集組件採集到的所有第二圖像採集組件。第二目標對象指的是需要進行活體檢測或目標識別的對象。In some embodiments, object detection is performed on the second initial image to obtain a second target object included in the second initial image. Wherein, the second initial image here refers to the second initial image corresponding to the selected final first initial image, rather than all the second image acquisition components acquired by the second image acquisition component. The second target object refers to an object that needs to be subjected to liveness detection or target recognition.

從最終第一初始圖像和第二初始圖像中，查找出匹配的一組第一目標對象和第二目標對象。其中，最終第一初始圖像中可能包含多個第一目標對象，最終第二初始圖像中也可能包含多個第二目標對象。如果直接對最終第一初始圖像和最終第二初始圖像進行活體檢測，那麼不同的目標對象之間可能存在干擾噪音，因此，先將查找到的第一目標對象和第二目標對象進行匹配，若匹配成功，則利用最終第一初始圖像得到包含查找出的第一目標對象的一幀待檢測圖像，以及利用第二初始圖像，得到包含查找出的第二目標對象的另一幀待檢測圖像。From the final first initial image and the second initial image, a set of matching first target objects and second target objects are found. The final first initial image may contain multiple first target objects, and the final second initial image may also contain multiple second target objects. If the final first initial image and the final second initial image are directly subjected to live detection, there may be interference noise between different target objects. Therefore, the found first target object and the second target object are first matched. , if the matching is successful, use the final first initial image to obtain a frame of the image to be detected containing the found first target object, and use the second initial image to obtain another image containing the found second target object Frame the image to be detected.

因此，通過先對若干張第一初始圖像進行目標檢測和跟蹤，選擇出符合條件的第一初始圖像，然後再對選擇出的第一初始圖像對應的第二初始圖像進行目標檢測，在一定程度上減輕了對系統資源的佔用。Therefore, by first performing target detection and tracking on several first initial images, a first initial image that meets the conditions is selected, and then target detection is performed on the second initial image corresponding to the selected first initial image. , to a certain extent, reduce the occupation of system resources.

在一些實施方式中，將最終第一初始圖像中的第一目標對象所在區域按照預設比例進行第一外擴，並提取第一外擴之後的區域作為一幀待檢測圖像。預設比例可以是原第一目標對象所在區域的上下左右各外擴二分之一，當然，在其他實施例中，也可外擴其他比例，例如上下左右各外擴三分之一，因此，關於預設比例此處不做限定。例如，外擴之後的區域應當包含完整的第一目標對象，若原本第一目標對象處於第一初始圖像的邊緣，外擴之後的區域可能會超出原本第一初始圖像的邊緣，那麼只保留未超出第一初始圖像的部分。當然在其他公開實施例中，若外擴之後超出第一初始圖像的邊緣，也可用第一初始圖像中非第一目標對象的背景區域對超出第一初始圖像邊緣的區域進行填充或利用預設圖元值對這部分區域進行填充。當然，經過外擴之後的區域，不包括其他第一目標對象的中心點，使得提取到的一幀待檢測圖像中僅包含一個完整的第一目標對象，減少其他目標對象對該第一目標對象的影響，使得活體檢測的準確度更高。In some embodiments, the region where the first target object is located in the final first initial image is first expanded according to a preset ratio, and the region after the first expansion is extracted as a frame of the image to be detected. The preset ratio can be one-half of the top, bottom, left, and right of the area where the original first target object is located. Of course, in other embodiments, other ratios can also be expanded, for example, the top, bottom, left, and right are expanded by one-third. Therefore, , the preset ratio is not limited here. For example, the expanded area should contain the complete first target object. If the original first target object is at the edge of the first initial image, the expanded area may exceed the edge of the original first initial image. The portion not exceeding the first initial image is preserved. Of course, in other disclosed embodiments, if the expansion exceeds the edge of the first initial image, the background area of the first initial image that is not the first target object can also be used to fill in or fill in the area beyond the edge of the first initial image. Fill this part of the area with the preset primitive values. Of course, the area after the expansion does not include the center points of other first target objects, so that the extracted frame of the image to be detected only contains a complete first target object, reducing the number of other target objects to the first target object. The influence of the object makes the accuracy of living detection higher.

將第二初始圖像中的第二目標對象所在區域按照預設比例進行第二外擴，並提取第二外擴之後的區域作為另一幀待檢測圖像。這裡的預設比例同上述預設比例，第二外擴的方式同上述第一外擴，因此，這裡不再贅述。A second expansion is performed on the region where the second target object is located in the second initial image according to a preset ratio, and the region after the second expansion is extracted as another frame of the image to be detected. The preset ratio here is the same as the above-mentioned preset ratio, and the manner of the second external expansion is the same as that of the above-mentioned first external expansion, so the details are not repeated here.

通過對第一初始圖像中的第一目標對象以及對應第二初始圖像中的第二目標對象進行外擴提取，從而減輕了其他目標對象對活體檢測結果的影響，提高了活體檢測的精度。By performing expansion extraction on the first target object in the first initial image and the second target object in the corresponding second initial image, the influence of other target objects on the living body detection result is reduced, and the accuracy of living body detection is improved .

一些公開實施例中，對至少兩幀待檢測圖像進行活體檢測，得到關於目標對象的活體檢測結果。通過利用兩個圖像採集組件對環境中的目標對象進行拍攝得到幀同步的兩張初始圖像，可以利用兩個初始圖像的視場差進行活體檢測，使得活體檢測的準確度更高。在一些實施方式中，對每幀待檢測圖像，利用待檢測圖像對應的拍攝模型匹配的活體檢測模型對待檢測圖像進行活體檢測。其中，活體檢測模型是樣本圖像訓練得到的，其中，活體檢測模型的樣本圖像是利用活體檢測模型匹配的拍攝模式拍攝得到的，樣本圖像包括對活體目標拍攝得到的活體樣本圖像和對假體目標拍攝得到的假體樣本圖像，其中，假體樣本圖像包括二維靜態圖像、二維動態圖像、三維模具中的至少一種。這裡的二維圖像包括靜態紙質圖像，例如各類紙質的列印圖片、照片、圖片剪影、照片剪影等，還可包括二維靜態非紙質圖像，例如各類材質的印染圖片等，還可包括二維靜態電子螢幕，例如手機、平板電腦、顯示器上顯示的動態視頻攻擊等，三維模具包括三維面具攻擊以及三維頭模攻擊等。在實施中活體檢測方式可以是，活體檢測模型獲取兩幀待檢測圖像的特徵，其中，這裡的特徵包括活體和非活體的特徵，然後給予提取到的特徵判斷待檢測圖像中的目標對象是否為活體，最後輸出檢測結果。其中，當待檢測圖像是圖像採集組件在第一拍攝模式下拍攝得到，那麼活體檢測模型則為第一拍攝模式對應的活體檢測模型，例如，第一拍攝模式為彩色拍攝模式，則活體檢測模型則為彩色活體檢測模型，當待檢測圖像是圖像採集組件在第二拍攝模式下拍攝得到，那麼活體檢測模型則為第二拍攝模式對應的活體檢測模型，例如，第二拍攝模式為紅外拍攝模式，則活體檢測模型則為紅外活體檢測模型。通過將拍攝模式與活體檢測模型對應，使得能夠更有針對性的對待檢測圖像進行活體檢測，使得活體檢測的準確度更高，同時，利用多樣的樣本對活體檢測模型進行訓練使得活體檢測模型的適用性更強，檢測結果的準確度更高。In some disclosed embodiments, in vivo detection is performed on at least two frames of images to be detected to obtain in vivo detection results about the target object. By using two image acquisition components to capture a target object in the environment to obtain two initial images with frame synchronization, the difference in the field of view of the two initial images can be used to perform living body detection, so that the accuracy of living body detection is higher. In some embodiments, for each frame of the image to be detected, the image to be detected is detected by using a living body detection model matched with the shooting model corresponding to the image to be detected. Wherein, the living body detection model is obtained by training a sample image, wherein the sample image of the living body detection model is obtained by using the shooting mode matched by the living body detection model, and the sample image includes the living body sample image obtained by shooting the living body target and the A prosthetic sample image obtained by photographing a prosthetic target, wherein the prosthetic sample image includes at least one of a two-dimensional static image, a two-dimensional dynamic image, and a three-dimensional mold. The two-dimensional images here include static paper images, such as printed pictures, photos, picture silhouettes, photo silhouettes, etc. of various types of paper, and can also include two-dimensional static non-paper images, such as printing and dyeing pictures of various materials, etc. It can also include two-dimensional static electronic screens, such as mobile phones, tablet computers, dynamic video attacks displayed on monitors, etc., and three-dimensional molds include three-dimensional mask attacks and three-dimensional head mold attacks. In the implementation, the living body detection method may be that the living body detection model obtains the features of two frames of images to be detected, wherein the features here include the features of living bodies and non-living bodies, and then the extracted features are used to determine the target object in the images to be detected. Whether it is a living body, and finally output the detection result. Wherein, when the image to be detected is captured by the image acquisition component in the first shooting mode, the living body detection model is the living body detection model corresponding to the first shooting mode. For example, if the first shooting mode is the color shooting mode, the living body detection model is The detection model is a color living body detection model. When the image to be detected is captured by the image acquisition component in the second shooting mode, the living body detection model is the living body detection model corresponding to the second shooting mode. For example, the second shooting mode Infrared shooting mode, the living body detection model is an infrared living body detection model. By correlating the shooting mode with the living body detection model, the living body detection can be performed in a more targeted manner on the images to be detected, so that the accuracy of the living body detection is higher. The applicability is stronger, and the accuracy of the detection results is higher.

一些公開實施例中，在得到關於目標對象的活體檢測結果之後，在活體檢測結果為目標對象屬於活體的情況下，對其中一幀待檢測圖像進行目標識別，得到目標對象的識別結果。在一些實施方式中，對通過第一初始圖像得到的一幀待檢測圖像進行目標識別，當然，在他實施例中，也可對通過第二初始圖像得到的一幀待檢測圖像進行目標識別。例如對其中一幀待檢測圖像進行特徵提取，得到關於目標對象的目標特徵。獲取目標特徵分別與至少一個預存特徵的相似度，基於相似度，確定目標對象的識別結果。預存特徵指的是預設目標的特徵資料，例如，當目標對象為人臉時，預設目標則也是人臉，預存特徵則是預設人臉的特徵資料。在一些實施方式中，將目標對象的目標特徵與資料庫中所有預設目標的特徵資料進行比對，輸出與各個預設目標的特徵資料的相似度，並進行相似度排名，判斷最高的相似度是否大於相似度閾值，若大於則認定識別成功，若不大於相似度閾值，則認定識別不成功。其中在識別結果為目標對象被識別成功的情況下，執行與目標對象的身份匹配的聯動控制。在一些實施方式中，在目標對象的身份屬於第一類身份的情況下，控制關聯的門體打開，和/或將目標對象的身份發送給關聯的第一通信設備，以使第一通信設備基於目標對象的身份進行第一類身份相關的業務。第一通信設備可以是後端系統，例如當相關的業務可以是做考勤、迎賓等等應用。而在目標對象的身份屬於第二類身份時，則控制外界設備發出警報，和/或將目標對象的身份發送給第一通信設備以使第一通信設備基於目標對象的身份進行第二類身份相關的業務。這裡的第二類身份相關的業務可以是佈防布控、軌跡檢索等應用。通過將識別結果執行聯動控制使得聯動過程更加方便，進一步地，通過目標識別結果聯動開門或聯動報警或其他相關業務，在一定程度上保障了通行的安全以及起到視頻監控的作用。In some disclosed embodiments, after obtaining the living body detection result about the target object, if the living body detection result is that the target object belongs to the living body, target recognition is performed on one of the images to be detected to obtain the target object recognition result. In some embodiments, target recognition is performed on a frame of images to be detected obtained from the first initial image. Of course, in other embodiments, a frame of images to be detected obtained from the second initial image can also be identified. Perform target recognition. For example, feature extraction is performed on one of the frames to be detected to obtain target features about the target object. The similarity between the target feature and at least one pre-stored feature is obtained, and based on the similarity, the recognition result of the target object is determined. The pre-stored feature refers to the feature data of the preset target. For example, when the target object is a face, the preset target is also a face, and the pre-stored feature is the feature data of the preset face. In some embodiments, the target feature of the target object is compared with the feature data of all preset targets in the database, the similarity with the feature data of each preset target is output, and the similarity is ranked to determine the highest similarity Whether the degree is greater than the similarity threshold, if it is greater than the similarity threshold, it is considered that the recognition is successful, and if it is not greater than the similarity threshold, the recognition is considered unsuccessful. When the recognition result is that the target object is successfully recognized, the linkage control matching the identity of the target object is performed. In some embodiments, when the identity of the target object belongs to the first type of identity, the associated door is controlled to open, and/or the identity of the target object is sent to the associated first communication device, so that the first communication device The first type of identity-related business is performed based on the identity of the target object. The first communication device may be a back-end system, for example, when the related business may be applications such as attendance, greeting and the like. When the identity of the target object belongs to the second type of identity, the external device is controlled to issue an alarm, and/or the identity of the target object is sent to the first communication device, so that the first communication device can perform the second type of identity based on the identity of the target object. related business. The second type of identity-related services here can be applications such as arming and control, and trajectory retrieval. The linkage process is more convenient by performing linkage control on the identification results. Further, through the linkage of target identification results to open doors or linkage alarms or other related services, to a certain extent, the safety of traffic is guaranteed and the role of video surveillance is played.

一些公開實施例中，在得到關於目標對象的活體檢測結果之後，若活體檢測結果為目標對象不屬於活體的情況下，發送關於活體檢測結果的第一通知。在一些實施方式中，將待檢測圖像和活體檢測結果中的至少一者進行第一編碼，並將第一編碼的結果打包至第一通知，以發送給第二通信設備。或者經過編碼之後的資料儲存到記憶體中。第二通信設備可以是與拍攝設備具有關聯關係的其他設備或其他系統。In some disclosed embodiments, after obtaining the living body detection result about the target object, if the living body detection result is that the target object does not belong to the living body, a first notification about the living body detection result is sent. In some embodiments, at least one of the image to be detected and the living body detection result is first encoded, and the result of the first encoding is packaged into a first notification to be sent to the second communication device. Or the encoded data is stored in the memory. The second communication device may be another device or other system having an associated relationship with the photographing device.

在得到目標對象的識別結果之後，在識別結果為目標對象未被識別成功的情況下，發送關於識別結果的第二通知。例如將待檢測圖像和識別結果中的至少一者進行第二編碼，並將第二編碼的結果打包至第二通知以發送給第三通信設備。其中，本發明實施例中提到的編碼可以是硬體編碼，也可以是軟體編碼。這裡的第三通信設備也是與拍攝設備具有關聯關係的設備或系統。其中，第一通信設備、第二通信設備以及第三通信設備可以是同一設備也可以不同的設備。After the recognition result of the target object is obtained, if the recognition result is that the target object has not been successfully recognized, a second notification about the recognition result is sent. For example, the second encoding is performed on at least one of the image to be detected and the recognition result, and the result of the second encoding is packaged into the second notification to be sent to the third communication device. The encoding mentioned in the embodiments of the present invention may be hardware encoding or software encoding. The third communication device here is also a device or system that has an associated relationship with the photographing device. The first communication device, the second communication device and the third communication device may be the same device or different devices.

通過只有在活體檢測結果為活體的情況下進行目標識別，減輕了後續目標識別的計算量，當目標對象不屬於活體時，將檢測結果以第一通知的方式發送出去以使得後續能夠對檢測結果進行記錄。By performing target recognition only when the living body detection result is a living body, the calculation amount of subsequent target recognition is reduced. When the target object does not belong to a living body, the detection result is sent in the form of a first notification, so that the subsequent detection results can be compared. record.

上述方案，通過拍攝設備所處環境的當前亮度資訊來調整圖像採集組件的拍攝模式，使得能夠在不同的光照環境下，選擇合適的圖像採集組件來獲取圖像，從而獲取到的圖像的品質更好。In the above solution, the shooting mode of the image acquisition component is adjusted by the current brightness information of the environment where the shooting device is located, so that an appropriate image acquisition component can be selected to acquire images under different lighting environments, so that the acquired images better quality.

其中，拍攝處理方法的執行主體可以是拍攝處理裝置或具有拍攝處理裝置的設備，例如，拍攝處理方法可以由終端設備或伺服器或其它處理設備執行，其中，終端設備可以為拍攝設備（例如，攝影機），使用者設備（User Equipment，UE）、移動設備、使用者終端、終端、蜂窩電話、無線電話、個人數位助理（Personal Digital Assistant，PDA）、手持設備、計算設備、車載設備、可穿戴設備等。在一些可能的實現方式中，該拍攝處理方法可以通過處理器調用記憶體中儲存的電腦可讀指令的方式來實現。The execution subject of the photographing processing method may be a photographing processing apparatus or a device having a photographing processing apparatus, for example, the photographing processing method may be executed by a terminal device or a server or other processing equipment, wherein the terminal device may be a photographing device (for example, Cameras), User Equipment (UE), Mobile Devices, User Terminals, Terminals, Cell Phones, Cordless Phones, Personal Digital Assistants (PDAs), handheld devices, computing devices, in-vehicle devices, wearables equipment, etc. In some possible implementations, the photographing processing method may be implemented by the processor calling computer-readable instructions stored in the memory.

下面結合一個示例對上述拍攝方法進行說明，然而值得注意的是，該示例僅是為了更好地說明本發明，並不構成對本發明的不當限定。The above-mentioned shooting method will be described below with reference to an example, however, it should be noted that this example is only for better illustrating the present invention, and does not constitute an improper limitation of the present invention.

基於雙目成像的活體檢測的技術和設備在門禁控制、人臉支付等單目標、環境光照較為簡單的場景的應用越來越多，但是在多目標、光照環境較為複雜的視頻監控、無感人臉通行場景則使用受限。The technology and equipment of live detection based on binocular imaging are more and more used in single-target and simple ambient lighting scenarios such as access control and face payment. The use of face traffic scenes is limited.

一方面相關的技術和設備往往是針對整幅圖像中的僅存在一張人臉的情況進行深度學習的訓練，對於畫面中有多個人臉的場景魯棒性低；另外一方面，目前的雙目活體檢測技術多為可見光+紅外的方式，在暗光場景下需要可見光和紅外分別補光，因此會造成較大的光污染，影響使用體驗，同時在室外強光場景，紅外成像會受到強紅外線干擾，影響成像品質，從而影響活體檢測結果；On the one hand, related technologies and equipment are often trained in deep learning for the situation where there is only one face in the whole image, and the robustness is low for scenes with multiple faces in the picture; on the other hand, the current The binocular live detection technology is mostly in the form of visible light + infrared. In dark light scenes, visible light and infrared light are required to be supplemented separately, so it will cause greater light pollution and affect the user experience. At the same time, in outdoor strong light scenes, infrared imaging will be affected by strong light. Infrared interference, affecting the imaging quality, thus affecting the results of live detection;

此外，當前人臉門禁一體機受限於雙目基線距離較小，對遠距離人臉目標的視差較小，只能近距離使用，無法兼顧視頻監控；而基於視頻監控的門禁系統均為單目攝影機，雙目攝影機無法在後端進行幀同步，因此無法進行有效的活體檢測，不能保證通行的安全性；以上兩種方法均無法兼顧視頻監控和通行安全，用戶如需對出入口進行門禁控制及視頻監控需要安裝視頻監控攝影機、人臉門禁設備，系統較為複雜整體成本較高；In addition, the current face access control all-in-one is limited by the small binocular baseline distance, and the parallax for long-distance face targets is small, so it can only be used at close range and cannot take into account video surveillance; while the access control systems based on video surveillance are all single The binocular camera cannot perform frame synchronization at the back end, so it cannot perform effective liveness detection and cannot guarantee the safety of passage; the above two methods cannot take into account video surveillance and passage safety. If the user needs to control the entrance and exit And video surveillance needs to install video surveillance cameras and face access control equipment, the system is more complex and the overall cost is high;

最後，當前人臉門禁系統僅支援白名單通行，無法進行黑名單報警；而人臉安防布控攝影機僅支援黑名單布控，無法聯動門禁控制。Finally, the current face access control system only supports whitelist access, and cannot perform blacklist alarm; while the face security deployment camera only supports blacklist deployment, and cannot be linked to access control.

基於此，本發明示例提供一種兼顧視頻監控及閘禁控制的雙目監控攝影機，解決複雜光照環境的監控場景下的多目標活體檢測和人臉識別報警和通行問題。利用基於深度學習的雙目可見光和雙目紅外活體檢測模型的切換更好的適應各種光照環境的監控場景，提高可用性並降低光污染。利用雙目大基線設計及外擴抓圖策略，實現大場景、遠目標、多目標的人臉活體檢測和識別，同時兼顧安防監控和門禁通行。利用單目檢測跟蹤，另一目僅做優選幀檢測的策略，降低系統計算資源佔用，提升系統效率。Based on this, an example of the present invention provides a binocular surveillance camera that takes into account both video surveillance and gate control, and solves the problem of multi-target living detection and face recognition alarm and passage in a surveillance scene with a complex lighting environment. The switching of binocular visible light and binocular infrared live detection models based on deep learning is used to better adapt to monitoring scenarios of various lighting environments, improve usability and reduce light pollution. Using the binocular large baseline design and the external expansion capture strategy, the face detection and recognition of large scenes, distant targets, and multi-targets are realized, while taking into account security monitoring and access control. The strategy of using monocular detection and tracking and the other only for optimal frame detection reduces the occupation of system computing resources and improves system efficiency.

圖3A是本發明示例提供的拍攝處理方法的系統方塊圖，如圖3A所示，包括圖像採集模組31、人臉檢測跟蹤模組32、活體檢測模組33、人臉識別模組34、編解碼模組35、聯動控制模組36以及網路傳輸模組37，可實現監控場景下的多目標人臉活體檢測和識別，並可聯動門禁系統進行門開關控制。3A is a system block diagram of a shooting processing method provided by an example of the present invention, as shown in FIG. 3A, including an image acquisition module 31, a face detection and tracking module 32, a living body detection module 33, and a face recognition module 34 , the codec module 35, the linkage control module 36 and the network transmission module 37, can realize the multi-target face detection and recognition in the monitoring scene, and can link the access control system to control the door switch.

其中，圖像採集模組31用於雙目同步採集圖像資料，人臉檢測跟蹤模組32用於對多目標人臉檢測跟蹤並獲取人臉圖片；活體檢測模組33用於判斷獲取到的人臉圖片是否為活體，如果判斷人臉圖片中的對象為活體，先通過人臉識別模組34進行人臉對比和屬性識別後，再通過編解碼模組35對人臉圖片進行編碼；如果判斷人臉圖片中的對象為非活體，直接通過編解碼模組35對人臉圖片進行編碼。人臉識別模組34輸出的識別結果發送到聯動控制模組36，在聯動控制模組36通過韋根/繼電器等聯動門開關、報警輸出、指示燈等。編解碼模組35、聯動控制模組36及網路傳輸模組37是在對各模組的輸出結果進行進一步處理的模組。Among them, the image acquisition module 31 is used for binocular synchronous acquisition of image data, the face detection and tracking module 32 is used for multi-target face detection and tracking and obtains face pictures; the living body detection module 33 is used for judging the acquired Whether the face picture is a living body, if it is judged that the object in the face picture is a living body, after performing face comparison and attribute recognition through the face recognition module 34, the face picture is encoded by the codec module 35; If it is determined that the object in the face picture is not a living body, the face picture is directly encoded by the encoding and decoding module 35 . The recognition result output by the face recognition module 34 is sent to the linkage control module 36, and the linkage control module 36 links the door switch, alarm output, indicator light, etc. through Wiegand/relay. The codec module 35 , the linkage control module 36 and the network transmission module 37 are modules that further process the output results of each module.

編解碼模組35對圖像採集模組31採集到的視頻、人臉檢測跟蹤模組32、活體檢測模組33、人臉識別模組34輸出的結果圖片資料進行編碼，對註冊入庫圖片進行解碼。The encoding and decoding module 35 encodes the video collected by the image acquisition module 31, the result picture data output by the face detection and tracking module 32, the living body detection module 33, and the face recognition module 34, and performs the registration and storage process on the pictures. decoding.

聯動控制模組36根據人臉識別結果進行外設的聯動。如識別結果在白名單庫中，則可以通過設置聯動韋根介面或繼電器1進行門禁控制；如識別結果在黑名單中，則可以通過設置聯動繼電器2進行報警設備控制，可外接聲光報警器等外接報警設備。The linkage control module 36 performs linkage of peripheral devices according to the face recognition result. If the recognition result is in the whitelist library, you can set the linkage Wiegand interface or relay 1 for access control; if the recognition result is in the blacklist, you can set the linkage relay 2 to control the alarm device, and an external sound and light alarm can be used. and other external alarm equipment.

網路傳輸模組37同時與編解碼模組35、聯動控制模組36之間通信連接，即時傳輸抓拍圖片、活體檢測結果、識別結果、視頻流、控制命令等資訊。網路傳輸模組37將視頻資料、圖片資料、各類解析結果等通過標準的網路通訊協定進行傳輸，實現與其他系統通訊。The network transmission module 37 communicates with the codec module 35 and the linkage control module 36 at the same time, and transmits information such as snapshot pictures, living body detection results, recognition results, video streams, and control commands in real time. The network transmission module 37 transmits video data, picture data, various analysis results, etc. through a standard network communication protocol, so as to realize communication with other systems.

在圖像採集模組31中主要為提升環境適應性、降低功耗和光污染做了以下設計。The following designs are mainly made in the image acquisition module 31 to improve environmental adaptability, reduce power consumption and light pollution.

一方面，雙目大基線設計：採集模組由兩顆彩色轉黑白的圖像感測器及對應的鏡頭、紅外截止濾光片組成，根據監控場景的不同，兩個圖像感測器的基線距離可設計為60毫米到150毫米。另一方面，雙目彩轉黑同步切換：兩顆圖像感測器可根據視頻亮度檢測或光敏裝置進行同步彩色轉黑白切換，並同時控制紅外截止濾光片的切換，保證雙目同時獲取穩定的彩色或黑白圖像。這樣，利用雙圖像感測器的設計和優化，加大雙目的視場差，解決監控場景下人臉距離遠時視差較小導致的活體精度問題，從而利用單台設備兼顧視頻監控和門禁控制。On the one hand, the binocular large baseline design: the acquisition module consists of two color-to-black and white image sensors, corresponding lenses, and infrared cut-off filters. The baseline distance can be designed from 60mm to 150mm. On the other hand, binocular color-to-black synchronous switching: two image sensors can perform synchronous color-to-black and white switching according to video brightness detection or photosensitive devices, and simultaneously control the switching of infrared cut-off filters to ensure simultaneous acquisition of binoculars Stable color or black and white images. In this way, the design and optimization of the dual image sensor is used to increase the binocular field of view, and to solve the problem of in vivo accuracy caused by the small parallax when the face is far away in the monitoring scene, so that a single device can be used for both video surveillance and access control. control.

再一方面，單補光燈設計：在攝影機切換為黑白模式時，紅外補光燈打開，為雙目同時補光；攝影機切換為彩色模式時，紅外燈關閉，降低能耗；因同步彩色/黑白切換，在照度較高時的彩色模式下，無需補光燈，降低能耗，並避免室外強光下對紅外補光的干擾，提升雙目活體識別的魯棒性；在照度較低時的黑白模式下，僅需紅外補光，避免可見光補光的光污染。On the other hand, the design of single fill light: when the camera is switched to black and white mode, the infrared fill light is turned on to fill light for both eyes at the same time; when the camera is switched to color mode, the infrared light is turned off to reduce energy consumption; Black and white switching, in the color mode when the illumination is high, there is no need for supplementary light, reducing energy consumption, avoiding the interference of infrared supplementary light under strong outdoor light, and improving the robustness of binocular living body recognition; when the illumination is low In the black and white mode, only infrared fill light is needed to avoid light pollution of visible light fill light.

這樣，通過對環境照度的判斷進行監控圖像的彩色黑白轉換，保證監控圖像品質的同時，同步切換雙目活體檢測算法，在彩色模式下使用雙目可見光活體演算法，在黑白模式使用雙目紅外活體演算法，去除可見光補光燈依賴和光污染。In this way, the color and black-and-white conversion of the monitoring image is performed by judging the environmental illumination, so as to ensure the quality of the monitoring image, and simultaneously switch the binocular in vivo detection algorithm. Infrared in vivo algorithm to remove visible light fill light dependence and light pollution.

圖3B是本發明示例提供的基於雙目幀同步採集圖像資料的流程示意圖，如圖3B所示，所述方法至少包括以下步驟：步驟S301b，判斷當前拍攝環境的照度；這裡，環境照度可以為亮度；步驟S302b，在照度高於閾值的情況下，雙目保持彩色模式採集圖像資料；步驟S303b，在照度低於閾值的情況下，雙目同步切換黑白模式並同步移除紅外截止濾光片，同時打開紅外燈採集圖像資料；步驟S304b，向活體檢測模組33發出保持載入雙目可見光（Red Green Blue，RGB）活體模型/切換雙目紅外（Near Infrared，NIR）活體模型的控制指令，向編碼模組35及人臉檢測跟蹤模組32發送圖像資料。FIG. 3B is a schematic flow chart of collecting image data based on binocular frame synchronization provided by an example of the present invention. As shown in FIG. 3B , the method includes at least the following steps: Step S301b, judging the illuminance of the current shooting environment; here, the illuminance of the environment can be brightness; Step S302b, in the case that the illuminance is higher than the threshold, the binocular maintains the color mode to collect image data; Step S303b, when the illuminance is lower than the threshold, binocular synchronously switches the black and white mode and synchronously removes the infrared cut-off filter, and simultaneously turns on the infrared lamp to collect image data; Step S304b, issue a control command to the living body detection module 33 to keep loading the binocular visible light (Red Green Blue, RGB) living body model/switch the binocular infrared (Near Infrared, NIR) living body model, and send it to the encoding module 35 and the human face. The detection and tracking module 32 sends image data.

本發明示例通過採集模組的大基線設計獲取監控場景下目標的較大視差，為活體檢測模型提供更多的特徵資訊，提升整個監控場景的活體精度。有效解決了相關技術中只能應用在2米之內的場景使用，無法兼顧監控和通行的問題。The example of the present invention obtains the large parallax of the target in the monitoring scene through the large baseline design of the acquisition module, provides more feature information for the living body detection model, and improves the living body accuracy of the entire monitoring scene. It effectively solves the problem that the related technologies can only be used in scenarios within 2 meters, and cannot take into account monitoring and traffic.

本發明示例的人臉檢測跟蹤模組32的輸入來自於圖像採集模組31的兩路嚴格幀同步的輸入，也就是說在送入人臉檢測跟蹤模組32之前對兩路圖像均需做預處理。其中，主攝影頭採集的圖像在人臉檢測跟蹤模組32中進行人臉檢測、目標跟蹤，獲取人臉檢測結果及關鍵點資訊，並對每一個跟蹤的目標進行編號，以在多目標場景中區分不同的人臉目標；對檢測跟蹤到的人臉進行多維度品質綜合判斷，從而根據品質判斷結果進行選幀操作，並對篩選人臉所在幀對應副攝影頭圖像幀進行人臉檢測，並對雙目中分別檢測到的人臉進行匹配，匹配成功後根據人臉檢測框按一定比例外擴後得到各自的人臉抓拍圖，並將兩張人臉抓拍圖及關鍵點資訊送入活體檢測模組33。The input of the face detection and tracking module 32 in the example of the present invention comes from the input of two channels of strict frame synchronization of the image acquisition module 31 , that is to say, before being sent to the face detection and tracking module 32 , the two channels of images are Preprocessing is required. Among them, the image collected by the main camera is used for face detection and target tracking in the face detection and tracking module 32, to obtain the face detection results and key point information, and to number each tracked target, so that the multi-target Different face targets are distinguished in the scene; multi-dimensional quality comprehensive judgment is carried out on the detected and tracked faces, so as to perform frame selection operation according to the quality judgment results, and the frame of the screened face corresponds to the sub-camera image frame. Detect and match the faces detected in the binoculars. After the matching is successful, the face detection frame is expanded in a certain proportion to obtain the respective face snapshots, and the two face snapshots and key point information are obtained. into the living body detection module 33 .

圖3C是本發明示例提供的人臉檢測跟蹤過程的流程示意圖，如圖3C所示，所述方法至少包括以下步驟。FIG. 3C is a schematic flowchart of a face detection and tracking process provided by an example of the present invention. As shown in FIG. 3C , the method includes at least the following steps.

步驟S301c，獲取主攝影頭待檢測視頻幀；其中，主攝影頭相當於第一圖像採集組件，待檢測視頻幀中的圖像相當於第一初始圖像。Step S301c, acquiring the video frame to be detected by the main camera; The main camera is equivalent to the first image acquisition component, and the image in the video frame to be detected is equivalent to the first initial image.

步驟S302c，判斷視頻幀中每一幀圖像中是否有人臉；如果當前幀圖像中有人臉（相當於目標對象），則執行步驟S303c；如果當前幀圖像中無人臉，則執行步驟S304c。Step S302c, judging whether there is a face in each frame image in the video frame; If there is a face (equivalent to the target object) in the current frame image, step S303c is executed; if there is no face in the current frame image, step S304c is executed.

步驟S303c，跟蹤檢測到的人臉。Step S303c, track the detected face.

步驟S304c，進行下一幀圖像的檢測；繼續按照步驟S302c判斷下一幀圖像中是否有人臉。Step S304c, detecting the next frame of image; Continue to judge whether there is a face in the next frame of image according to step S302c.

步驟S305c，選取符合活體檢測要求的圖片；基於多維度圖像品質對檢測的人臉圖像進行選幀操作。其中，圖像品質判斷包括單不僅限於以下維度：置信度、人臉角度、模糊度、人臉圖元大小、綜合品質分數等，並可單獨以某個維度作為品質篩選條件，通過人臉品質判斷的篩選，可以減少送入活體及識別模組的資料，降低系統資源消耗，並降低低品質圖片輸入對活體檢測、人臉識別的精度影響。Step S305c, select a picture that meets the requirements of living body detection; The frame selection operation is performed on the detected face image based on the multi-dimensional image quality. Among them, the image quality judgment includes but is not limited to the following dimensions: confidence, face angle, ambiguity, face primitive size, comprehensive quality score, etc., and a certain dimension can be used as a quality screening condition. Judgment screening can reduce the data sent to the living body and the recognition module, reduce system resource consumption, and reduce the impact of low-quality image input on the accuracy of living body detection and face recognition.

步驟S306c，通過同步的幀號獲取副攝影頭對應幀圖像；其中，副攝影頭相當於第二圖像採集組件，副攝影頭對應幀圖像相當於第二初始圖像。本發明示例在副攝影頭的人臉檢測上，區別於其它方案的副攝影頭每一幀均進行人臉檢測的策略，本發明在主攝影頭的檢測跟蹤結果通過品質篩選後，才對副攝影頭的對應幀進行人臉檢測，並進行人臉匹配，以區分同一幀的多個人臉目標，在確保雙目抓取的人臉滿足活體檢測要求（主副攝影頭輸出的為同一個人臉、圖像品質較高）的同時，降低對系統資源的佔用。Step S306c, obtaining the frame image corresponding to the sub-camera through the synchronized frame number; Wherein, the sub-camera is equivalent to the second image acquisition component, and the frame image corresponding to the sub-camera is equivalent to the second initial image. The example of the present invention is in the face detection of the sub-camera, which is different from the strategy of performing face detection for each frame of the sub-camera in other schemes. The corresponding frame of the camera is subjected to face detection, and face matching is performed to distinguish multiple face targets in the same frame, so as to ensure that the face captured by the binocular meets the requirements of living body detection (the output of the main and auxiliary cameras is the same face. , higher image quality), while reducing the occupation of system resources.

步驟S307c，對副攝影頭對應幀圖像進行人臉檢測。Step S307c, performing face detection on the frame image corresponding to the sub-camera.

步驟S308c，將主副攝影頭檢出的人臉進行匹配。In step S308c, the faces detected by the main and sub cameras are matched.

步驟S309c，基於匹配成功的雙目人臉圖片得到兩張人臉抓拍圖，送入活體檢測模組33。其中，人臉抓拍圖相當於待檢測圖像。In step S309c, two face snapshots are obtained based on the successfully matched binocular face pictures, and sent to the living body detection module 33. Among them, the face snapshot image is equivalent to the image to be detected.

在主攝影頭檢測到人臉並通過品質篩選之後，在人臉檢測框的基礎上進行一定比例的外擴並抓取圖片，保證整個人臉均可出現在圖片中，同理，副攝影頭的對應人臉圖片也同樣比例外擴抓圖。此步操作可基本杜絕送入活體檢測的畫面中出現多個目標，從而降低畫面中同時存在活體和非活體時對活體檢測結果的影響，保證活體精度的同時提高通行效率。這樣，基於深度學習的人臉檢測技術在整幅圖像中截取人臉區域進行活體檢測，解決多人臉目標不能同時進行活體檢測的問題。After the main camera detects the face and passes the quality screening, a certain proportion of the face detection frame is expanded and the picture is captured to ensure that the entire face can appear in the picture. Similarly, the sub camera The corresponding face picture of the same scale is also expanded. This step can basically prevent multiple targets from appearing in the screen sent to the live body detection, thereby reducing the impact on the live body detection results when there are live bodies and non-live bodies in the screen at the same time, ensuring the accuracy of the live body and improving the traffic efficiency. In this way, the face detection technology based on deep learning intercepts the face region in the entire image for liveness detection, which solves the problem that multiple face targets cannot be simultaneously detected for liveness.

本發明示例通過主攝影頭做檢測跟蹤，並進行品質優選，之後再根據優選結果所在幀對應副攝影頭的圖像幀進行人臉檢測，避免雙目同時逐幀檢測時佔用雙倍檢測計算資源造成的系統資源浪費。同時針對相關技術中為全圖入活體檢測，無法解決畫面中多人臉的活體檢測問題，通過人臉外擴抓圖的策略，降低同一張圖片中同時出現多張人臉且同時有活體及非活體的情況，降低干擾，對檢測人臉的活體檢測更準確，可適用於多人臉目標場景。In the example of the present invention, the main camera is used for detection and tracking, and the quality is optimized, and then face detection is performed according to the image frame of the sub-camera corresponding to the frame where the optimized result is located, so as to avoid double detection computing resources when binocular simultaneous frame-by-frame detection is used. waste of system resources. At the same time, in view of the fact that the whole image is input into the living body detection in the related technology, the problem of living body detection of multiple faces in the picture cannot be solved. Through the strategy of face expansion and capture, the simultaneous appearance of multiple faces in the same picture and the simultaneous presence of living bodies and In the case of non-living bodies, the interference is reduced, and the live body detection for detecting faces is more accurate, which can be applied to scenes with multiple faces.

圖3D是本發明示例提供的對獲取的人臉圖片進行活體檢測的流程示意圖，如圖3D所示，所述方法至少包括以下步驟。FIG. 3D is a schematic flowchart of performing live detection on an acquired face picture provided by an example of the present invention. As shown in FIG. 3D , the method includes at least the following steps.

步驟S301d，根據照度判斷切換雙目可見光或雙目紅外模型；其中，照度可以通過亮度表徵，雙目可見光或雙目紅外模型為與待檢測圖像對應的拍攝模式匹配的活體檢測模組。雙目可見光活體模型：在照度較高時，使用雙目可見光活體模型，此模型輸入為雙目分別同步採集的對應圖像幀的可見光人臉圖片，輸出是否活體的判斷結果。此模型是通過大量安防場景下雙目彩色模式拍攝的人臉資料、各類非活體攻擊的資料訓練而來。雙目紅外活體模型：在照度較低時，使用雙目紅外活體模型，此模型輸入為雙目分別同步採集的對應圖像幀的紅外人臉圖片，輸出是否活體的判斷結果。此模型是通過大量安防場景下雙目黑白模式拍攝的人臉資料、各類非活體攻擊的資料訓練而來。模型自動切換：模型的切換由採集模組的照度判斷自動進行切換，無需人工作業。照度高時採集模組切換為彩色模式的同時，活體檢測模組切換雙目可見光活體模型；照度低時採集模組切換黑白模式的同時，活體檢測模組切換雙目紅外活體模型。這樣，利用基於深度學習技術的雙目可見光和雙目紅外的活體檢測算法去除可見光補光依賴，解決可見光污染問題及紅外干擾問題，提升用戶體驗和實際效果。Step S301d, switch the binocular visible light or binocular infrared model according to the illuminance judgment; The illuminance can be characterized by brightness, and the binocular visible light or binocular infrared model is a living body detection module matching the shooting mode corresponding to the image to be detected. Binocular visible light living body model: When the illumination is high, the binocular visible light living body model is used. The input of this model is the visible light face picture of the corresponding image frame collected by the binocular synchronously, and the output is the judgment result of whether it is living body. This model is trained through a large number of face data captured in binocular color mode in security scenes and data of various non-living attacks. Binocular infrared living body model: When the illumination is low, the binocular infrared living body model is used. The input of this model is the infrared face pictures of the corresponding image frames collected by the binoculars synchronously, and the judgment result of whether it is living body is output. This model is trained through a large number of face data shot in binocular black-and-white mode in security scenes and data of various non-living attacks. Automatic model switching: The model switching is automatically switched by the illumination judgment of the acquisition module, without manual work. When the illumination is high, the acquisition module switches to the color mode, and the living detection module switches the binocular visible light living model; when the illumination is low, the acquisition module switches to the black and white mode, and the living detection module switches the binocular infrared living model. In this way, the binocular visible light and binocular infrared live detection algorithm based on deep learning technology is used to remove the dependence of visible light supplementary light, solve the problem of visible light pollution and infrared interference, and improve the user experience and actual effect.

步驟S302d，人臉檢測跟蹤模組32輸出的一對人臉圖像。Step S302d, a pair of face images output by the face detection and tracking module 32 is detected.

步驟S303d，活體檢測模組33判斷人臉圖像中的人臉是否為活體；如果判斷結果是人臉為活體，執行步驟S304d；如果判斷結果是人臉為非活體，執行步驟S305d。Step S303d, the living body detection module 33 judges whether the human face in the human face image is a living body; If the determination result is that the human face is a living body, step S304d is performed; if the determination result is that the human face is a non-living body, step S305d is performed.

步驟S304d，主攝影頭人臉圖像送入人臉識別模組34。In step S304d, the face image of the main camera is sent to the face recognition module 34.

步驟S305d，所有人臉圖像及活體判斷結果送入編碼模組35和網路傳輸模組36。In step S305d, all face images and living body judgment results are sent to the encoding module 35 and the network transmission module 36.

在活體檢測模組33的實現中，利用雙目攝影機的視場差，可通過深度學習網路獲取更多的非活體圖像資訊。因需要在全天候的安防場景下使用，白天夜晚光照環境的差異較大，且因監控距離較遠，如夜間採用白光需要較大功率，造成光污染且會對通行人員的視覺感官造成不適；而白天室外，如採用可見光+紅外的方式，則會因為強光反射造成紅外圖像過曝，影響室外的精度和使用效果。因此在本發明示例中，根據安防場景的特點，在白天（照度高）雙目同時為彩色模式，並採用雙目可見光輸入的活體模型，排除白天對紅外圖像的干擾；在夜晚（照度低）時雙目自動切換至黑白模式，並切換為雙目紅外輸入的活體模型，無需白光補光，解決光污染的問題。In the implementation of the living body detection module 33, more non-living body image information can be obtained through the deep learning network by using the field of view difference of the binocular camera. Because it needs to be used in all-weather security scenarios, the lighting environment during the day and night is quite different, and because the monitoring distance is long, such as using white light at night requires high power, causing light pollution and discomfort to the visual senses of passers-by; Outdoors, if the visible light + infrared method is used, the infrared image will be overexposed due to strong light reflection, which will affect the accuracy and use effect outdoors. Therefore, in the example of the present invention, according to the characteristics of the security scene, in the daytime (high illuminance) binoculars are in color mode at the same time, and a living model with binocular visible light input is adopted to eliminate the interference of infrared images during the day; in the night (low illuminance) ), the binocular automatically switches to black and white mode, and switches to a living model with binocular infrared input, eliminating the need for white light to fill the light and solving the problem of light pollution.

經過針對性設計和訓練的雙目可見光活體模型及雙目紅外活體模型，在安防場景下可精確的防止的非活體攻擊包括：二維靜態紙質圖像例如各類紙質的列印圖片、照片、圖片剪影、照片剪影等；二維靜態非紙質圖像例如各類材質的印染圖片等；二維靜態電子螢幕，包括手機、平板電腦、顯示器上顯示的靜態圖片攻擊；二維動態電子螢幕，包括手機、平板電腦、顯示器上顯示的動態視頻攻擊；三維面具攻擊；三維頭模攻擊。The binocular visible light living model and binocular infrared living model that have been designed and trained in a targeted manner can accurately prevent non-living attacks in security scenarios including: two-dimensional static paper images such as various types of paper printed pictures, photos, Picture silhouettes, photo silhouettes, etc; Dynamic video attack displayed on mobile phones, tablet computers, monitors; 3D mask attack; 3D head model attack.

由通過上述攻擊資料訓練好的深度神經網路對雙目攝影機採集到的的一對待檢測圖像分別進行綜合特徵提取。深度神經網路基於提取的綜合特徵中判斷活體待檢測圖像是否活體人臉，得到待檢測圖像是否為活體的檢測結果。The deep neural network trained by the above-mentioned attack data separately performs comprehensive feature extraction on the images to be detected collected by the binocular camera. Based on the extracted comprehensive features, the deep neural network judges whether the image to be detected in the living body is a living human face, and obtains the detection result of whether the image to be detected is a living body.

相關技術中主要為可見光+紅外雙目活體檢測，在暗光場景需要可見光補光燈提升目標亮度，造成光污染，對用戶不夠友好、無感，在強光環境下，環境光會對紅外攝影頭的成像有較大干擾，影響活體精度。本發明示例通過雙目可見光和雙目紅外的活體模型切換，配合雙目的彩轉黑和單紅外燈設計，有效降低光污染、減少強環境光對紅外燈的干擾、提升各種光照環境的活體檢測的魯棒性。The related technologies are mainly visible light + infrared binocular live detection. In dark light scenes, visible light supplementary lights are required to improve the brightness of the target, causing light pollution, which is not user-friendly and insensitive. In a strong light environment, the ambient light will affect the infrared camera head. The imaging has large interference, which affects the accuracy of the living body. The example of the present invention can effectively reduce light pollution, reduce the interference of strong ambient light on infrared lamps, and improve the performance of living body detection in various lighting environments by switching between binocular visible light and binocular infrared living body models, combined with binocular color-to-black and single-infrared lamp design. robustness.

人臉識別模組34在接收到通過活體檢測的人臉圖像後，通過已訓練的深度神經網路提取人臉特徵資料，並與已註冊提取特徵的人臉特徵庫進行特徵比對，輸出相似度。對相似度進行排名，取相似度最高的對應人臉特徵庫中的特徵（簡稱底庫特徵）再與設定的識別閾值進行比較，高於閾值的判斷為同一個人，觸發下一步動作。After receiving the face image detected by the living body, the face recognition module 34 extracts the face feature data through the trained deep neural network, and compares the features with the registered face feature database for extracting features, and outputs the result. similarity. The similarity is ranked, and the features in the corresponding face feature library with the highest similarity (referred to as the base library feature) are compared with the set recognition threshold. If the threshold is higher than the threshold, it is judged as the same person, and the next action is triggered.

圖3E是本發明示例提供的人臉識別過程的流程示意圖，如圖3E所示，至少包括以下步驟。FIG. 3E is a schematic flowchart of a face recognition process provided by an example of the present invention. As shown in FIG. 3E , at least the following steps are included.

步驟S301e，獲取通過活體檢測的主攝影頭人臉圖像。Step S301e, acquiring a face image of the main camera detected by the living body.

步驟S302e，對主攝影頭人臉圖像進行人臉特徵提取。Step S302e, performing facial feature extraction on the facial image of the main camera.

步驟S303e，與已註冊的人臉特徵庫的特徵進行比對，確定相似度最高的底庫特徵；人臉識別的人臉特徵庫分為兩大類，白名單和黑名單，分別進行不同的業務應用和聯動進一步的動作：白名單可聯動門禁開門，並將識別資訊推送到後端系統做考勤、迎賓等應用；黑名單可聯動報警，並將識別資訊推送到後端系統做佈防布控、軌跡檢索等應用。從而對黑名單人員進行黑名單報警提示，對白名單人員進行基於韋根協議或繼電器的門禁控制。Step S303e, compare with the features of the registered face feature library, and determine the bottom library feature with the highest similarity; The facial feature database of face recognition is divided into two categories, whitelist and blacklist, which are used for different business applications and further actions. Welcome and other applications; the blacklist can be linked to the alarm, and the identification information can be pushed to the back-end system for deployment and control, trajectory retrieval and other applications. In this way, the blacklisted personnel are alerted to the blacklist, and the access control based on the Wiegand protocol or relay is performed for the whitelisted personnel.

步驟S304e，判斷相似度是否超過識別閾值；如果相似度超過識別閾值，則執行步驟S305e；如果相似度未超過識別閾值，則執行步驟S306e。Step S304e, judging whether the similarity exceeds the recognition threshold; If the similarity exceeds the identification threshold, execute step S305e; if the similarity does not exceed the identification threshold, execute step S306e.

步驟S305e，根據特徵庫類型，設置聯動下一步動作。Step S305e, according to the type of the feature library, set the next action of linkage.

步驟S306e，將人臉圖像及相似度結果送入編碼模組36和網路傳輸模組37。In step S306e, the face image and the similarity result are sent to the encoding module 36 and the network transmission module 37.

在本發明示例中，人臉識別模組34可對通過活體檢測的可信人臉進行人臉特徵提取，並與已建模入庫的人臉圖片進行比對，用於確認人員身份並聯動進一步的動作，包括開門、報警、資訊推送等。針對不同的人臉底庫類型採用不同的業務策略，對黑名單人員進行報警推送和聯動指定繼電器；對白名單人員可通過韋根協議或指定繼電器聯動門禁控制。可同時兼顧安防布控和人臉通行。In the example of the present invention, the face recognition module 34 can perform face feature extraction on the credible face detected by the living body, and compare it with the face image that has been modeled and stored, so as to confirm the identity of the person and link further Actions, including door opening, alarm, information push, etc. Different business strategies are adopted for different face base database types, and the blacklisted personnel can be alarmed and linked to designated relays; the whitelisted personnel can be linked with access control through Wiegand protocol or designated relays. It can take into account security deployment and face access at the same time.

本發明示例可以應用在需要快速通行的門禁控制場景，可以利用本發明在監控視場下的多目標人臉檢測、活體檢測和人臉識別實現無感通行，無需通行人員主動配合。在同時需要門禁控制和視頻監控的場景，可以利用本發明示例使用一台設備兼顧兩類需求。在夜間使用中，可以利用本發明的無可見光補光設計，杜絕可見光補光對使用者的視覺刺激，提升使用者體驗。在白天戶外使用中，可解決相關方案紅外攝影頭受強光干擾過曝而影響通行、活體檢測效率的問題，提升可用性。在考勤應用中，可對人員進行無感考勤，避免現在主動式打卡方式下員工忘記打卡的問題。The example of the present invention can be applied to the access control scene requiring fast passage, and the multi-target face detection, living body detection and face recognition of the present invention in the monitoring field of view can be used to realize non-sensing passage without the active cooperation of passers-by. In a scenario where access control and video surveillance are required at the same time, an example of the present invention can be used to satisfy both types of requirements with one device. In use at night, the non-visible light supplementary light design of the present invention can be used to prevent the visual stimulation of the user by the visible light supplementary light and improve the user experience. In the daytime outdoor use, it can solve the problem that the infrared camera of the related solution is interfered by strong light and overexposed, which affects the efficiency of passage and living body detection, and improves the usability. In the attendance application, it is possible to perform non-inductive attendance for personnel to avoid the problem of employees forgetting to punch in under the current active punch-in method.

請參閱圖4，圖4是本發明實施例拍攝處理裝置的結構示意圖。拍攝處理裝置40包括亮度獲取模組41以及模式切換模組42。亮度獲取模組41配置為獲取拍攝設備所處環境的當前亮度資訊，其中，拍攝設備包括至少兩個圖像採集組件；模式切換模組42配置為將至少一個所述圖像採集組件調整為與所述當前亮度資訊匹配的拍攝模式，其中，不同所述拍攝模式是採用不同的光形成圖像。Please refer to FIG. 4 , which is a schematic structural diagram of a photographing processing apparatus according to an embodiment of the present invention. The photographing processing device 40 includes a brightness acquiring module 41 and a mode switching module 42 . The brightness acquisition module 41 is configured to acquire the current brightness information of the environment where the photographing device is located, wherein the photographing device includes at least two image acquisition components; the mode switching module 42 is configured to adjust at least one of the image acquisition components to be the same as the image acquisition component. The current brightness information matches the shooting modes, wherein different shooting modes use different lights to form images.

一些公開實施例中，模式切換模組42將至少一個圖像採集組件調整為與當前亮度資訊匹配的拍攝模式，包括：在當前亮度資訊大於亮度閾值的情況下，將至少一個圖像採集組件調整為第一拍攝模式；在當前亮度資訊不大於亮度閾值的情況下，將至少一個圖像採集組件調整為第二拍攝模式。In some disclosed embodiments, the mode switching module 42 adjusts the at least one image capture component to a shooting mode that matches the current brightness information, including: when the current brightness information is greater than the brightness threshold, adjusting the at least one image capture component. is the first shooting mode; when the current brightness information is not greater than the brightness threshold, adjust at least one image acquisition component to the second shooting mode.

上述方案，利用亮度資訊是否大於閾值的條件來調整圖像採集組件的拍攝模式，使得獲取到的圖像的品質更好。In the above solution, the condition of whether the brightness information is greater than the threshold is used to adjust the shooting mode of the image acquisition component, so that the quality of the acquired image is better.

一些公開實施例中，第一拍攝模式為彩色拍攝模式，第二拍攝模式為紅外拍攝模式。In some disclosed embodiments, the first shooting mode is a color shooting mode, and the second shooting mode is an infrared shooting mode.

一些公開實施例中，拍攝設備中的每個圖像採集組件的拍攝模式相同。In some disclosed embodiments, the photographing mode of each image acquisition component in the photographing device is the same.

上述方案，通過在亮度資訊大於閾值時，選擇彩色拍攝模式，當亮度資訊小於閾值時，選擇紅外拍攝模式，使得白天亮度較高紅外線過強時，減少紅外成像會受到強紅外線的干擾，而在夜晚亮度較暗時，選擇紅外拍攝模式，減輕了對可見光的依賴，從而提高了成像品質。In the above scheme, when the brightness information is greater than the threshold value, the color shooting mode is selected, and when the brightness information is less than the threshold value, the infrared shooting mode is selected, so that when the brightness is high during the day and the infrared rays are too strong, the infrared imaging will be reduced by the interference of strong infrared rays. When the brightness is dark at night, the infrared shooting mode is selected to reduce the dependence on visible light, thereby improving the imaging quality.

一些公開實施例中，在第一拍攝模式為彩色拍攝模式，第二拍攝模式為紅外拍攝模式的情況下，模式切換模組42將至少一個圖像採集組件調整為第一拍攝模式，包括以下至少一個步驟：關閉拍攝設備的紅外補光燈；將拍攝設備的紅外截止濾光片移動至需調整為第一拍攝模式的圖像採集組件的入光通道中；將至少一個圖像採集組件調整為第二拍攝模式，包括以下至少一個步驟：打開拍攝設備的紅外補光燈；將拍攝設備的紅外截止濾光片移動至需調整為第二拍攝模式的圖像採集組件的入光通道以外。In some disclosed embodiments, when the first shooting mode is the color shooting mode and the second shooting mode is the infrared shooting mode, the mode switching module 42 adjusts the at least one image capturing component to the first shooting mode, including at least the following: One step: turn off the infrared fill light of the shooting device; move the infrared cut-off filter of the shooting device to the light incident channel of the image acquisition component that needs to be adjusted to the first shooting mode; adjust at least one image acquisition component to The second shooting mode includes at least one of the following steps: turning on the infrared fill light of the shooting device; moving the infrared cut-off filter of the shooting device outside the light incident channel of the image capture component to be adjusted to the second shooting mode.

上述方案，在圖像採集組件為彩色拍攝模式時，說明環境亮度較高，紅外光比較強，因此，無需打開紅外補光燈，且將紅外截止濾光片放置在圖像採集組件的入光通道內，對紅外光進行過濾，減輕了紅外光對拍攝到的圖像的影響，而在圖像採集組件為紅外拍攝模式時，說明環境亮度較暗，才利用打開紅外補光燈，使得拍攝到的圖像更清楚即品質更好。In the above solution, when the image acquisition component is in color shooting mode, it means that the ambient brightness is high and the infrared light is relatively strong. Therefore, it is not necessary to turn on the infrared fill light, and the infrared cut-off filter is placed on the incoming light of the image acquisition component. In the channel, the infrared light is filtered to reduce the influence of infrared light on the captured image. When the image acquisition component is in the infrared shooting mode, it means that the ambient brightness is dark, and the infrared fill light is turned on to make the shooting. The resulting image is clearer and of better quality.

一些公開實施例中，拍攝處理裝置40還包括圖像獲取模組、活體檢測模組（圖未示）。其中，圖像獲取模組配置為獲取至少兩幀初始圖像，其中，至少兩幀初始圖像為至少圖像採集組件對所處環境中的目標對象分別拍攝得到的；基於每幀初始圖像，對應得到包含目標對象的待檢測圖像；活體檢測模組，配置為對至少兩幀待檢測圖像進行活體檢測，得到關於目標對象的活體檢測結果。In some disclosed embodiments, the photographing processing device 40 further includes an image acquisition module and a living body detection module (not shown). Wherein, the image acquisition module is configured to acquire at least two frames of initial images, wherein the at least two frames of initial images are obtained by at least the image acquisition component respectively shooting the target object in the environment; , correspondingly obtains an image to be detected including the target object; the living body detection module is configured to perform living body detection on at least two frames of the to-be-detected image to obtain a living body detection result about the target object.

上述方案，通過利用拍攝設備的至少兩個圖像採集組件拍攝的圖像而獲取到的待檢測圖像用於活體檢測相對於單攝影頭採集到的圖像進行活體檢測，前者在活體檢測過程中可以利用兩張圖像之間的視場差，從而提高活體檢測的準確性。In the above solution, the to-be-detected image obtained by using images captured by at least two image acquisition components of the photographing device is used for living body detection. Compared with the image collected by a single camera, the former is used for living body detection during the living body detection process. The difference in the field of view between the two images can be exploited to improve the accuracy of liveness detection.

一些公開實施例中，至少兩個圖像採集組件包括第一圖像採集組件和第二圖像採集組件；圖像獲取模組基於每幀初始圖像，對應得到包含目標對象的待檢測圖像，包括：從第一圖像採集組件採集到的若干幀第一初始圖像中，選出最終第一初始圖像；獲取第二圖像採集組件採集的與最終第一初始圖像對應幀的第二初始圖像；分別利用最終第一初始圖像和第二初始圖像，對應得到包含目標對象的兩幀待檢測圖像。In some disclosed embodiments, the at least two image acquisition components include a first image acquisition component and a second image acquisition component; the image acquisition module correspondingly obtains an image to be detected containing the target object based on each frame of the initial image. , comprising: selecting a final first initial image from several frames of first initial images collected by the first image collection component; acquiring the first initial image of the frame corresponding to the final first initial image collected by the second image collection component Two initial images; using the final first initial image and the second initial image respectively, correspondingly obtain two frames of images to be detected including the target object.

上述方案，通過先在第一圖像採集組件採集到的若干張第一初始圖像，然後再利用選擇出的第一初始圖像以及對應的第二初始圖像獲得待檢測圖像，加快了選擇的速率，在一定程度上提高了獲取待檢測圖像的效率，而且相比於對兩個圖像採集組件的初始圖像分別進行選擇，本方案只需對其中一個圖像採集組件的初始圖像進行選擇，可以減少對處理資源的使用。In the above solution, by first obtaining several first initial images collected by the first image collecting component, and then using the selected first initial image and the corresponding second initial image to obtain the image to be detected, the speed of the process is accelerated. The selected rate improves the efficiency of acquiring the image to be detected to a certain extent, and compared with the selection of the initial images of the two image acquisition components, this scheme only needs to Image selection can reduce the use of processing resources.

一些公開實施例中，圖像獲取模組從第一圖像採集組件對目標對象拍攝得到的若干幀第一初始圖像中，選出最終第一初始圖像，包括：對每幀第一初始圖像進行目標檢測和跟蹤，得到每幀第一初始圖像所包含的第一目標對象；從若干幀第一初始圖像中，選擇第一目標對象符合活體檢測要求的第一初始圖像，作為最終第一初始圖像。In some disclosed embodiments, the image acquisition module selects a final first initial image from several frames of first initial images captured by the first image acquisition component of the target object, including: for each frame of the first initial image; Perform target detection and tracking to obtain the first target object included in the first initial image of each frame; from several frames of the first initial image, select the first initial image of the first target object that meets the requirements of living body detection, as Final first initial image.

一些公開實施例中，影像處理模組分別利用最終第一初始圖像和第二初始圖像，對應得到包含目標對象的兩幀待檢測圖像，包括：對第二初始圖像進行目標檢測，得到第二初始圖像所包含的第二目標對象；從最終第一初始圖像和第二初始圖像中，查找出匹配的一組第一目標對象和第二目標對象；利用最終第一初始圖像，得到包含查找出的第一目標對象的一幀待檢測圖像，以及利用第二初始圖像，得到包含查找出的第二目標對象的另一幀待檢測圖像。In some disclosed embodiments, the image processing module uses the final first initial image and the second initial image respectively to obtain two frames of images to be detected including the target object, including: performing target detection on the second initial image, Obtain the second target object contained in the second initial image; find out a set of matching first target objects and second target objects from the final first initial image and the second initial image; use the final first initial image image to obtain a frame of images to be detected including the first target object found, and obtain another frame of images to be detected including the second target object found by using the second initial image.

上述方案，通過將先對若干張第一初始圖像進行目標檢測和跟蹤，選擇出符合條件的第一初始圖像，然後再對選擇出的第一初始圖像對應的第二初始圖像進行目標檢測，在一定程度上減輕了對系統資源的佔用。The above scheme, by first performing target detection and tracking on several first initial images, selecting a first initial image that meets the conditions, and then performing a second initial image corresponding to the selected first initial image. Target detection reduces the occupation of system resources to a certain extent.

一些公開實施例中，圖像獲取模組從若干幀第一初始圖像中，選擇第一目標對象符合活體檢測要求的第一初始圖像，作為最終第一初始圖像，包括：對於每幀第一初始圖像：基於第一目標對象的至少一個品質因數，得到第一目標對象的品質分數；其中，第一目標對象的品質因數包括以下至少一種：第一目標對象的置信度、角度、大小、模糊度以及第一目標對象所在的第一初始圖像的模糊度；選擇第一目標對象的品質分數大於預設分數閾值的第一初始圖像，作為最終第一初始圖像；影像處理模組43利用最終第一初始圖像，得到包含查找出的第一目標對象的一幀待檢測圖像，包括：將最終第一初始圖像中的第一目標對象所在區域按照預設比例進行第一外擴，並提取第一外擴之後的區域作為一幀待檢測圖像；影像處理模組43利用第二初始圖像，得到包含查找出的第二目標對象的另一幀待檢測圖像，包括：將第二初始圖像中的第二目標對象所在區域按照預設比例進行第二外擴，並提取第二外擴之後的區域作為另一幀待檢測圖像。In some disclosed embodiments, the image acquisition module selects, from several frames of first initial images, the first initial image of the first target object that meets the requirements of living body detection as the final first initial image, including: for each frame; The first initial image: obtaining a quality score of the first target object based on at least one quality factor of the first target object; wherein, the quality factor of the first target object includes at least one of the following: a confidence level of the first target object, an angle, a size, blur degree, and blur degree of the first initial image where the first target object is located; select the first initial image with the quality score of the first target object greater than the preset score threshold as the final first initial image; image processing The module 43 uses the final first initial image to obtain a frame of the image to be detected that includes the first target object found, including: performing the final first initial image in the area where the first target object is located according to a preset ratio. The first expansion is performed, and the area after the first expansion is extracted as a frame of the image to be detected; the image processing module 43 uses the second initial image to obtain another frame of the image to be detected that includes the found second target object The image includes: performing a second expansion of the area where the second target object is located in the second initial image according to a preset ratio, and extracting the area after the second expansion as another frame of the image to be detected.

上述方案，通過獲取品質分數滿足條件的第一初始圖像以及第一初始圖像對應的第二初始圖像，減輕了外界因素對活體檢測的影響，從而使得得到的待檢測圖像的活體檢測結果越準確；其中，通過對第一初始圖像中的第一目標對象以及對應第二初始圖像中的第二目標對象進行外擴提取，從而減輕了其他目標對象對活體檢測結果的影響，提高了活體檢測的精度。In the above solution, by obtaining the first initial image whose quality score satisfies the condition and the second initial image corresponding to the first initial image, the influence of external factors on the detection of living bodies is alleviated, so that the obtained images to be detected can be detected by living bodies. The more accurate the result is; wherein, by performing expansion extraction on the first target object in the first initial image and the second target object in the corresponding second initial image, the influence of other target objects on the living body detection result is reduced, Improves the accuracy of liveness detection.

一些公開實施例中，活體檢測模組對至少兩幀待檢測圖像進行活體檢測，得到關於目標對象的活體檢測結果，包括：對於每幀待檢測圖像，利用待檢測圖像對應的拍攝模式匹配的活體檢測模型對待檢測圖像進行活體檢測。In some disclosed embodiments, the in vivo detection module performs in vivo detection on at least two frames of images to be detected, and obtains in vivo detection results about the target object, including: for each frame of the to-be-detected image, using the shooting mode corresponding to the to-be-detected image. The matched liveness detection model performs liveness detection on the image to be detected.

上述方案，通過將拍攝模式與活體檢測模型對應，使得能夠更有針對性的對待檢測圖像進行活體檢測，使得活體檢測的準確度更高。In the above solution, by correlating the shooting mode with the living body detection model, it is possible to perform the living body detection on the image to be detected in a more targeted manner, so that the accuracy of the living body detection is higher.

一些公開實施例中，活體檢測模型是樣本圖像訓練得到的；其中，活體檢測模型的樣本圖像是利用活體檢測模型匹配的拍攝模式拍攝得到的；樣本圖像包括對活體目標拍攝得到的活體樣本圖像和對假體目標拍攝得到的假體樣本圖像，假體目標包括二維靜態圖像、二維動態圖像、三維模具中的至少一種。In some disclosed embodiments, the living body detection model is obtained by training a sample image; wherein, the sample image of the living body detection model is obtained by using a shooting mode matched by the living body detection model; the sample image includes a living body obtained by photographing a living body target. A sample image and a prosthetic sample image obtained by photographing a prosthetic target, the prosthetic target includes at least one of a two-dimensional static image, a two-dimensional dynamic image, and a three-dimensional mold.

上述方案，利用多樣的樣本對活體檢測模型進行訓練使得活體檢測模型的適用性更強，檢測結果的準確度更高。In the above solution, using a variety of samples to train the living body detection model makes the living body detection model more applicable and the detection result more accurate.

一些公開實施例中，拍攝處理裝置40包括目標識別模組（圖未示），通信模組（圖未示），在活體檢測模組對至少兩幀待檢測圖像進行活體檢測，得到關於目標對象的活體檢測結果之後，包括以下至少一步：在活體檢測結果為目標對象屬於活體的情況下，目標識別模組對其中一幀待檢測圖像進行目標識別，得到目標對象的識別結果；在活體檢測結果為目標對象不屬於活體的情況下，通信模組發送關於活體檢測結果的第一通知。In some disclosed embodiments, the photographing processing device 40 includes a target recognition module (not shown) and a communication module (not shown). After the living body detection result of the object, it includes the following at least one step: when the living body detection result is that the target object belongs to the living body, the target recognition module performs target recognition on one of the images to be detected, and obtains the target object recognition result; When the detection result is that the target object does not belong to the living body, the communication module sends a first notification about the living body detection result.

上述方案，只有在活體檢測結果為活體的情況下進行目標識別，減輕了後續目標識別的計算量，當目標對象不屬於活體時，將檢測結果以第一通知的方式發送出去以使得後續能夠對檢測結果進行記錄。In the above scheme, target recognition is only performed when the living body detection result is a living body, which reduces the calculation amount of subsequent target recognition. When the target object does not belong to a living body, the detection result is sent as a first notification so that subsequent Record the test results.

一些公開實施例中，目標識別模組對其中一幀待檢測圖像進行目標識別，得到目標對象的識別結果，包括：對其中一幀待檢測圖像進行特徵提取，得到關於目標對象的目標特徵；獲取目標特徵分別與至少一個預存特徵的相似度；基於相似度，確定目標對象的識別結果。In some disclosed embodiments, the target recognition module performs target recognition on one of the frames to be detected to obtain the recognition result of the target object, including: performing feature extraction on one of the frames to be detected to obtain target features about the target object. ; obtain the similarity between the target feature and at least one pre-stored feature respectively; and determine the recognition result of the target object based on the similarity.

上述方案，通過相似度的比對來確定識別結果，使得目標識別結果更有依據且準確。In the above scheme, the recognition result is determined by comparing the similarity, so that the target recognition result is more basis and accurate.

一些公開實施例中，拍攝處理裝置40包括聯動模組（圖未示），在得到目標對象的識別結果之後，包括：在識別結果為目標對象被識別成功的情況下，聯動模組執行與目標對象的身份匹配的聯動控制；在識別結果為目標對象未被識別成功的情況下，通信模組發送關於識別結果的第二通知。In some disclosed embodiments, the photographing processing device 40 includes a linkage module (not shown in the figure), and after obtaining the recognition result of the target object, it includes: if the recognition result is that the target object is successfully identified, the linkage module executes the operation with the target object. Linkage control of object identity matching; in the case that the recognition result is that the target object has not been successfully recognized, the communication module sends a second notification about the recognition result.

上述方案，通過將識別結果執行聯動控制使得聯動過程更加方便。The above solution makes the linkage process more convenient by performing linkage control on the identification result.

一些公開實施例中，聯動模組執行與目標對象的身份匹配的聯動控制，包括：在目標對象的身份屬於第一類身份，則控制關聯的門體打開，和/或，將目標對象的身份發送給關聯的第一通信設備，以使第一通信設備基於目標對象的身份進行與第一類身份相關的業務；在目標對象的身份屬於第二類身份，則控制外接設備發出警報，和/或，通信模組將目標對象的身份發送給第一通信設備，以使第一通信設備基於目標對象的身份進行與第二類身份相關的業務；通信模組發送關於活體檢測結果的第一通知，包括：將待檢測圖像和活體檢測結果中的至少一者進行第一編碼，並將第一編碼的結果打包至第一通知以發送給第二通信設備；通信模組發送關於識別結果的第二通知，包括：將待檢測圖像和識別結果中的至少一者進行第二編碼，並將第二編碼的結果打包至第二通知以發送給第三通信設備。In some disclosed embodiments, the linkage module performs linkage control matching the identity of the target object, including: when the identity of the target object belongs to the first type of identity, controlling the opening of the associated door, and/or changing the identity of the target object. sent to the associated first communication device, so that the first communication device conducts a business related to the first type of identity based on the identity of the target object; when the identity of the target object belongs to the second type of identity, the external device is controlled to issue an alarm, and/ Or, the communication module sends the identity of the target object to the first communication device, so that the first communication device performs a business related to the second type of identity based on the identity of the target object; the communication module sends a first notification about the living body detection result , including: first encoding at least one of the image to be detected and the living body detection result, and packaging the result of the first encoding into a first notification to send to the second communication device; the communication module sends a message about the recognition result The second notification includes: performing second encoding on at least one of the image to be detected and the recognition result, and packaging the second encoding result into the second notification for sending to the third communication device.

上述方案，通過目標識別結果聯動開門或聯動報警或其他相關業務，在一定程度上保障了通行的安全以及起到視頻監控的作用。The above scheme, through the target recognition results, can be linked to open the door or linked to an alarm or other related services, to a certain extent, to ensure the safety of traffic and to play a role in video surveillance.

請參閱圖5，圖5是本發明實施例電子設備的結構示意圖。電子設備50包括記憶體52和處理器51，處理器51配置為執行記憶體52中儲存的程式指令，以實現上述任一拍攝處理方法實施例中的步驟。在一個實施場景中，電子設備50可以包括但不限於：拍攝設備、微型電腦、伺服器，此外，電子設備50還可以包括筆記型電腦、平板電腦等移動設備，在此不做限定。Please refer to FIG. 5 , which is a schematic structural diagram of an electronic device according to an embodiment of the present invention. The electronic device 50 includes a memory 52 and a processor 51, and the processor 51 is configured to execute program instructions stored in the memory 52 to implement the steps in any of the above-mentioned embodiments of the shooting processing method. In an implementation scenario, the electronic device 50 may include, but is not limited to, a photographing device, a microcomputer, and a server. In addition, the electronic device 50 may also include a mobile device such as a notebook computer and a tablet computer, which is not limited herein.

處理器51配置為控制其自身以及記憶體52以實現上述任一拍攝處理方法實施例中的步驟。處理器51還可以稱為中央處理單元（Central Processing Unit，CPU）。處理器51可能是一種積體電路晶片，具有信號的處理能力。處理器51還可以是通用處理器、數位訊號處理器（Digital Signal Processor, DSP）、專用積體電路（Application Specific Integrated Circuit, ASIC）、現場可程式設計閘陣列（Field-Programmable Gate Array, FPGA）或者其他可程式設計邏輯器件、分立門或者電晶體邏輯器件、分立硬體組件。通用處理器可以是微處理器或者該處理器也可以是任何常規的處理器等。另外，處理器51可以由積體電路晶片共同實現。The processor 51 is configured to control itself and the memory 52 to implement the steps in any of the above-mentioned embodiments of the shooting processing method. The processor 51 may also be referred to as a central processing unit (Central Processing Unit, CPU). The processor 51 may be an integrated circuit chip with signal processing capability. The processor 51 may also be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), or a field-programmable gate array (FPGA) Or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. In addition, the processor 51 may be commonly implemented by an integrated circuit die.

一些公開實施例中，電子設備50為拍攝設備50，拍攝設備50還包括至少兩個圖像採集組件53，其中，處理器51連接圖像採集組件53，以控制圖像採集組件53的拍攝模式。其中，兩個圖像採集組件53之間的基線距離大於預設距離閾值。例如基線距離為60毫米至150毫米。當然，在其他實施例中，基線的距離還可以更長，這主要取決於拍攝設備50的大小以及實際場景中的需求。In some disclosed embodiments, the electronic device 50 is a photographing device 50 , and the photographing device 50 further includes at least two image acquisition components 53 , wherein the processor 51 is connected to the image acquisition components 53 to control the photographing mode of the image acquisition components 53 . . Wherein, the baseline distance between the two image acquisition components 53 is greater than the preset distance threshold. For example, the baseline distance is 60 mm to 150 mm. Of course, in other embodiments, the distance of the baseline may be longer, which mainly depends on the size of the photographing device 50 and the requirements in the actual scene.

上述方案，通過設置基線距離在60毫米至150毫米，使得拍攝設備50的視程能夠達到8米左右，從而能夠兼顧監控和通行。In the above solution, by setting the baseline distance between 60 mm and 150 mm, the visual range of the photographing device 50 can reach about 8 meters, so that both monitoring and traffic can be taken into account.

一些公開實施例中，拍攝設備50還包括編解碼器54，配置為對待發送給外部通信設備的第一資料進行編碼，以及對外部通信設備發送的第二資料進行解碼。其中，編解碼器54還能對圖像採集組件53獲取到的圖像資料以及上述方法實施例中從若干幀第一初始圖像和第二初始圖像中選擇出來的最終第一初始圖像和第二初始圖像以及經過活體檢測之後的活體檢測結果以及待檢測圖像、經過目標識別之後的識別結果以及待檢測圖像進行編碼以使得這些資料能夠在後續進行儲存或傳輸。In some disclosed embodiments, the photographing device 50 further includes a codec 54 configured to encode the first material to be sent to the external communication device, and to decode the second material to be sent by the external communication device. Wherein, the codec 54 can also collect the image data obtained by the image acquisition component 53 and the final first initial image selected from several frames of the first initial image and the second initial image in the above method embodiment. The second initial image, the living body detection result after living body detection and the to-be-detected image, the recognition result after target recognition and the to-be-detected image are encoded so that these data can be stored or transmitted later.

上述方案，通過硬體進行編解碼則不需要佔用CPU，CPU就可以如釋重負，輕鬆上陣，承擔更多的其他任務。In the above scheme, encoding and decoding through hardware does not need to occupy the CPU, and the CPU can be relieved, easy to go into battle, and undertake more other tasks.

一些公開實施例中，拍攝設備50還包括聯動電路55，配置為向外部關聯設備發送聯動控制指令。聯動電路55可以是繼電器電路、韋根介面等In some disclosed embodiments, the photographing device 50 further includes a linkage circuit 55 configured to send a linkage control instruction to an external associated device. The linkage circuit 55 can be a relay circuit, a Wiegand interface, etc.

上述方案，通過將設置聯動電路55，建立了拍攝設備50和其他關聯設備的連接。In the above solution, by setting the linkage circuit 55, the connection between the photographing device 50 and other related devices is established.

一些公開實施例中，拍攝設備50還包括通信電路56，配置為與外部設備進行通信。In some disclosed embodiments, the camera device 50 further includes a communication circuit 56 configured to communicate with external devices.

上述方案，通過設置通信電路使得拍攝設備50更加智慧，使得資料的傳輸更加方便。In the above solution, by setting the communication circuit, the photographing device 50 is made more intelligent, and the data transmission is more convenient.

請參閱圖6，圖6是本發明實施例電腦可讀儲存介質的結構示意圖。電腦可讀儲存介質60儲存有能夠被處理器運行的程式指令61，程式指令61用於實現上述任一拍攝處理方法實施例中的步驟。電腦可讀儲存介質60可以是可以保存和儲存由指令執行設備使用的指令的有形設備，可為易失性儲存介質或非易失性儲存介質。Please refer to FIG. 6 , which is a schematic structural diagram of a computer-readable storage medium according to an embodiment of the present invention. The computer-readable storage medium 60 stores program instructions 61 that can be executed by the processor, and the program instructions 61 are used to implement the steps in any of the above-mentioned embodiments of the shooting processing method. Computer-readable storage medium 60 may be a tangible device that may hold and store instructions for use by the instruction execution device, and may be a volatile storage medium or a non-volatile storage medium.

在一些實施例中，本發明實施例提供的裝置具有的功能或包含的模組可以配置為執行上文方法實施例描述的方法，其實施過程可以參照上文方法實施例的描述。In some embodiments, the functions or modules included in the apparatus provided in the embodiments of the present invention may be configured to execute the methods described in the above method embodiments, and the implementation process may refer to the descriptions in the above method embodiments.

上文對各個實施例的描述傾向於強調各個實施例之間的不同之處，其相同或相似之處可以互相參考。The above description of various embodiments has tended to emphasize the differences between the various embodiments, the same or similarities may be referred to each other.

在本發明實施例所提供的幾個實施例中，應該理解到，所揭露的方法和裝置，可以通過其它的方式實現。例如，以上所描述的裝置實施方式僅僅是示意性的，例如，模組或單元的劃分，僅僅為一種邏輯功能劃分，實際實現時可以有另外的劃分方式，例如單元或組件可以結合或者可以集成到另一個系統，或一些特徵可以忽略，或不執行。另一點，所顯示或討論的相互之間的耦合或直接耦合或通信連接可以是通過一些介面，裝置或單元的間接耦合或通信連接，可以是電性、機械或其它的形式。In the several embodiments provided by the embodiments of the present invention, it should be understood that the disclosed methods and apparatuses may be implemented in other manners. For example, the device implementations described above are only illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other divisions. For example, units or components may be combined or integrated. to another system, or some features can be ignored, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.

另外，在本發明各個實施例中的各功能單元可以集成在一個處理單元中，也可以是各個單元單獨物理存在，也可以兩個或兩個以上單元集成在一個單元中。上述集成的單元既可以採用硬體的形式實現，也可以採用軟體功能單元的形式實現。集成的單元如果以軟體功能單元的形式實現並作為獨立的產品銷售或使用時，可以儲存在一個電腦可讀取儲存介質中。基於這樣的理解，本發明實施例的技術方案本質上或者說對現有技術做出貢獻的部分或者該技術方案的全部或部分可以以軟體產品的形式體現出來，該電腦軟體產品儲存在一個儲存介質中，包括若干指令用以使得一台電腦設備（可以是個人電腦，伺服器，或者網路設備等）或處理器（processor）執行本發明實施例各個實施方式方法的全部或部分步驟。而前述的儲存介質可以是但不局限於：U盤、移動硬碟、唯讀記憶體（ROM，Read-Only Memory）、隨機存取記憶體（RAM，Random Access Memory）、磁碟或者光碟等各種可以儲存程式碼的介質。In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware, or can be implemented in the form of software functional units. The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present invention are essentially or contribute to the prior art, or all or part of the technical solutions can be embodied in the form of software products, and the computer software products are stored in a storage medium. , including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (processor) to execute all or part of the steps of the various implementation methods of the embodiments of the present invention. The aforementioned storage medium can be, but is not limited to: U disk, removable hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or CD, etc. Various media that can store code.

工業實用性本發明實施例中，通過獲取拍攝設備所處環境的當前亮度資訊，其中，拍攝設備包括至少兩個圖像採集組件；將至少一個圖像採集組件調整為與當前亮度資訊匹配的拍攝模式；其中，不同拍攝模式是採用不同的光形成圖像。上述方案，通過拍攝設備所處環境的當前亮度資訊來調整圖像採集組件的拍攝模式，使得能夠在不同的光照環境下，選擇合適的拍攝模式來採集得到圖像，以使採集到的圖像的品質更好。能夠在一定程度上提升後續活體檢測的準確度。Industrial Applicability In the embodiment of the present invention, the current brightness information of the environment where the shooting device is located is obtained, wherein the shooting device includes at least two image acquisition components; the at least one image acquisition component is adjusted to a shooting mode matching the current brightness information; wherein , Different shooting modes use different light to form images. In the above solution, the shooting mode of the image acquisition component is adjusted by the current brightness information of the environment where the shooting device is located, so that an appropriate shooting mode can be selected to collect images under different lighting environments, so that the collected images better quality. It can improve the accuracy of subsequent live detection to a certain extent.

20:拍攝設備 21:第一圖像採集組件 22:第二圖像採集組件 31:圖像採集模組 32:人臉檢測跟蹤模組 33:活體檢測模組 34:人臉識別模組 35:編解碼模組 36:聯動控制模組 37:網路傳輸模組 40:拍攝處理裝置 41:亮度獲取模組 42:模式切換模組 50:電子設備 51:處理器 52:記憶體 53:圖像採集組件 54:編解碼器 55:聯動電路 56:通信電路 60:電腦可讀儲存介質 61:程式指令 S11~S12,S301b~S304b,S301c~S309c, S301d~S305d,S301e~S306e:步驟20: Shooting equipment 21: The first image acquisition component 22: Second image acquisition component 31: Image acquisition module 32: Face detection and tracking module 33: Liveness detection module 34: Face recognition module 35: Codec module 36: Linkage control module 37: Network transmission module 40: Shooting processing device 41: Brightness acquisition module 42: Mode switching module 50: Electronics 51: Processor 52: memory 53: Image acquisition components 54: Codec 55: Linkage circuit 56: Communication circuit 60: Computer-readable storage medium 61: Program command S11~S12, S301b~S304b, S301c~S309c, S301d~S305d, S301e~S306e: Steps

此處的附圖被併入說明書中並構成本說明書的一部分，這些附圖示出了符合本發明的實施例，並與說明書一起用於說明本發明實施例的技術方案。圖1是本發明實施例提供的拍攝處理方法的流程示意圖；圖2是本發明實施例提供的拍攝處理方法中具有兩圖像採集組件的拍攝設備的結構示意圖；圖3A是本發明示例提供的拍攝處理方法的系統方塊圖；圖3B是本發明示例提供的基於雙目幀同步採集圖像資料的流程示意圖；圖3C是本發明示例提供的人臉檢測跟蹤過程的流程示意圖；圖3D是本發明示例提供的對獲取的人臉圖片進行活體檢測的流程示意圖；圖3E是本發明示例提供的人臉識別過程的流程示意圖；圖4是本發明實施例提供的拍攝處理裝置的結構示意圖；圖5是本發明實施例提供的電子設備的結構示意圖；圖6是本發明實施例提供的電腦可讀儲存介質的結構示意圖。The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, and these drawings illustrate embodiments consistent with the present invention, and together with the description, serve to explain the technical solutions of the embodiments of the present invention. 1 is a schematic flowchart of a shooting processing method provided by an embodiment of the present invention; 2 is a schematic structural diagram of a photographing device having two image acquisition components in a photographing processing method provided by an embodiment of the present invention; 3A is a system block diagram of a shooting processing method provided by an example of the present invention; 3B is a schematic flow chart of collecting image data based on binocular frame synchronization provided by an example of the present invention; 3C is a schematic flowchart of a face detection and tracking process provided by an example of the present invention; 3D is a schematic flowchart of performing live detection on an acquired face picture provided by an example of the present invention; 3E is a schematic flowchart of a face recognition process provided by an example of the present invention; 4 is a schematic structural diagram of a photographing processing apparatus provided by an embodiment of the present invention; 5 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention; FIG. 6 is a schematic structural diagram of a computer-readable storage medium provided by an embodiment of the present invention.

S11~S12:步驟S11~S12: Steps

Claims

一種拍攝處理方法，包括：獲取拍攝設備所處環境的當前亮度資訊，其中，所述拍攝設備包括至少兩個圖像採集組件；將至少一個所述圖像採集組件調整為與所述當前亮度資訊匹配的拍攝模式；其中，不同所述拍攝模式是採用不同的光形成圖像。A shooting processing method, comprising: Acquiring current brightness information of the environment where the photographing device is located, wherein the photographing device includes at least two image acquisition components; adjusting at least one of the image capture components to a shooting mode that matches the current brightness information; Wherein, different shooting modes use different light to form images.

根據請求項1所述的方法，其中，所述將至少一個所述圖像採集組件調整為與所述當前亮度資訊匹配的拍攝模式，包括：在所述當前亮度資訊大於亮度閾值的情況下，將至少一個所述圖像採集組件調整為第一拍攝模式；在所述當前亮度資訊不大於所述亮度閾值的情況下，將至少一個所述圖像採集組件調整為第二拍攝模式。The method according to claim 1, wherein the adjusting at least one of the image capturing components to a shooting mode matching the current brightness information comprises: When the current brightness information is greater than the brightness threshold, adjusting at least one of the image capture components to a first shooting mode; Under the condition that the current brightness information is not greater than the brightness threshold, at least one of the image capturing components is adjusted to a second shooting mode.

根據請求項2所述的方法，其中，所述第一拍攝模式為彩色拍攝模式，所述第二拍攝模式為紅外拍攝模式。The method according to claim 2, wherein the first shooting mode is a color shooting mode, and the second shooting mode is an infrared shooting mode.

根據請求項1至3任一項所述的方法，其中，所述拍攝設備中的每個所述圖像採集組件的拍攝模式相同。The method according to any one of claims 1 to 3, wherein the shooting modes of each of the image capturing components in the shooting device are the same.

根據請求項2或3所述的方法，其中，在所述第一拍攝模式為彩色拍攝模式，所述第二拍攝模式為紅外拍攝模式的情況下，所述將至少一個所述圖像採集組件調整為第一拍攝模式，包括以下至少一個步驟：關閉所述拍攝設備的紅外補光燈；將所述拍攝設備的紅外截止濾光片移動至需調整為所述第一拍攝模式的圖像採集組件的入光通道中；所述將至少一個所述圖像採集組件調整為第二拍攝模式，包括以下至少一個步驟：打開所述拍攝設備的紅外補光燈；將所述拍攝設備的紅外截止濾光片移動至需調整為所述第二拍攝模式的圖像採集組件的入光通道以外。The method according to claim 2 or 3, wherein, when the first shooting mode is a color shooting mode and the second shooting mode is an infrared shooting mode, the at least one image acquisition component is Adjusting to the first shooting mode includes at least one of the following steps: Turn off the infrared fill light of the photographing device; moving the infrared cut-off filter of the photographing device to the light incident channel of the image acquisition component that needs to be adjusted to the first photographing mode; The adjusting at least one of the image capturing components to the second shooting mode includes at least one of the following steps: Turn on the infrared fill light of the shooting device; Move the infrared cut-off filter of the photographing device out of the light incident channel of the image acquisition component that needs to be adjusted to the second photographing mode.

根據請求項1至3任一項所述的方法，還包括：獲取至少兩幀初始圖像，其中，所述至少兩幀初始圖像為所述至少兩個圖像採集組件對所處環境中的目標對象分別拍攝得到的；基於每幀所述初始圖像，對應得到包含所述目標對象的待檢測圖像；對至少兩幀所述待檢測圖像進行活體檢測，得到關於所述目標對象的活體檢測結果。The method according to any one of claims 1 to 3, further comprising: acquiring at least two frames of initial images, wherein the at least two frames of initial images are obtained by respectively photographing the target object in the environment by the at least two image acquisition components; Based on the initial image of each frame, correspondingly obtain the to-be-detected image containing the target object; In vivo detection is performed on at least two frames of the to-be-detected images to obtain in vivo detection results about the target object.

根據請求項6所述的方法，其中，所述至少兩個圖像採集組件包括第一圖像採集組件和第二圖像採集組件；所述基於每幀所述初始圖像，對應得到包含所述目標對象的待檢測圖像，包括：從所述第一圖像採集組件採集到的若干幀第一初始圖像中，選出最終第一初始圖像；獲取所述第二圖像採集組件採集的與所述最終第一初始圖像對應幀的第二初始圖像；分別利用所述最終第一初始圖像和所述第二初始圖像，對應得到包含所述目標對象的兩幀待檢測圖像。The method according to claim 6, wherein the at least two image acquisition components include a first image acquisition component and a second image acquisition component; and the corresponding acquisition including the The image to be detected of the target object, including: Selecting a final first initial image from several frames of first initial images collected by the first image acquisition component; acquiring a second initial image of a frame corresponding to the final first initial image collected by the second image capture component; Using the final first initial image and the second initial image respectively, two frames of to-be-detected images containing the target object are correspondingly obtained.

根據請求項7所述的方法，其中，所述從所述第一圖像採集組件採集到的若干幀第一初始圖像中，選出最終第一初始圖像，包括：對每幀所述第一初始圖像進行目標檢測和跟蹤，得到每幀所述第一初始圖像所包含的第一目標對象；從所述若干幀所述第一初始圖像中，選擇所述第一目標對象符合活體檢測要求的所述第一初始圖像，作為所述最終第一初始圖像。The method according to claim 7, wherein the selecting the final first initial image from the several frames of the first initial images collected by the first image acquisition component comprises: Performing target detection and tracking on the first initial image of each frame to obtain the first target object included in the first initial image of each frame; From the several frames of the first initial images, the first initial images for which the first target object meets the requirements of living body detection are selected as the final first initial images.

根據請求項8所述的方法，其中，所述分別利用所述最終第一初始圖像和所述第二初始圖像，對應得到包含所述目標對象的兩幀待檢測圖像，包括：對所述第二初始圖像進行目標檢測，得到所述第二初始圖像所包含的第二目標對象；從所述最終第一初始圖像和所述第二初始圖像中，查找出匹配的一組所述第一目標對象和所述第二目標對象；利用所述最終第一初始圖像，得到包含查找出的所述第一目標對象的一幀待檢測圖像，以及利用所述第二初始圖像，得到包含查找出的所述第二目標對象的另一幀待檢測圖像。The method according to claim 8, wherein, using the final first initial image and the second initial image respectively to obtain two frames of images to be detected including the target object, comprising: performing target detection on the second initial image to obtain a second target object included in the second initial image; From the final first initial image and the second initial image, find out a matching set of the first target object and the second target object; Using the final first initial image to obtain a frame of images to be detected including the found first target object, and using the second initial image to obtain the second target object found Another frame of the image to be detected.

根據請求項9所述的方法，其中，所述從所述若干幀所述第一初始圖像中，選擇所述第一目標對象符合活體檢測要求的所述第一初始圖像，作為所述最終第一初始圖像，包括：對於每幀所述第一初始圖像：基於所述第一目標對象的至少一個品質因數，得到所述第一目標對象的品質分數；其中，所述第一目標對象的品質因數包括以下至少一種：所述第一目標對象的置信度、角度、大小、模糊度以及所述第一目標對象所在的第一初始圖像的模糊度；選擇所述第一目標對象的品質分數大於預設分數閾值的第一初始圖像，作為所述最終第一初始圖像；所述利用所述最終第一初始圖像，得到包含查找出的所述第一目標對象的一幀待檢測圖像，包括：將所述最終第一初始圖像中的所述第一目標對象所在區域按照預設比例進行第一外擴，並提取第一外擴之後的區域作為所述一幀待檢測圖像；所述利用所述第二初始圖像，得到包含查找出的所述第二目標對象的另一幀待檢測圖像，包括：將所述第二初始圖像中的所述第二目標對象所在區域按照預設比例進行第二外擴，並提取第二外擴之後的區域作為所述另一幀待檢測圖像。The method according to claim 9, wherein, from the several frames of the first initial images, the first initial image of the first target object that meets the requirements of living body detection is selected as the The final first initial image, including: For each frame of the first initial image: obtain a quality score of the first target object based on at least one quality factor of the first target object; wherein the quality factor of the first target object includes at least one of the following : the confidence, angle, size, and ambiguity of the first target object and the ambiguity of the first initial image where the first target object is located; Selecting a first initial image with a quality score of the first target object greater than a preset score threshold as the final first initial image; The use of the final first initial image to obtain a frame of images to be detected that includes the found first target object, including: performing a first expansion according to a preset ratio in the area where the first target object is located in the final first initial image, and extracting the area after the first expansion as the one frame of the image to be detected; Using the second initial image to obtain another frame of the image to be detected that includes the found second target object, including: The second expansion is performed on the area where the second target object is located in the second initial image according to a preset ratio, and the area after the second expansion is extracted as the other frame of the image to be detected.

根據請求項6至10任一項所述的方法，其中，所述對至少兩幀所述待檢測圖像進行活體檢測，得到關於所述目標對象的活體檢測結果，包括：對於每幀所述待檢測圖像，利用所述待檢測圖像對應的拍攝模式匹配的活體檢測模型對所述待檢測圖像進行活體檢測。The method according to any one of claims 6 to 10, wherein the performing in vivo detection on at least two frames of the to-be-detected images to obtain the in vivo detection result about the target object, comprising: For each frame of the to-be-detected image, the to-be-detected image is subjected to in vivo detection by using a living body detection model matched with a shooting mode corresponding to the to-be-detected image.

根據請求項11所述的方法，其中，所述活體檢測模型是樣本圖像訓練得到的；其中，所述活體檢測模型的樣本圖像是利用所述活體檢測模型匹配的拍攝模式拍攝得到的；所述樣本圖像包括對活體目標拍攝得到的活體樣本圖像和對假體目標拍攝得到的假體樣本圖像，所述假體目標包括二維靜態圖像、二維動態圖像、三維模具中的至少一種。The method according to claim 11, wherein the living detection model is obtained by training sample images; The sample image of the living body detection model is obtained by using the shooting mode matched by the living body detection model; the sample image includes the living body sample image obtained by photographing the living body target and the image obtained by photographing the prosthetic target. The prosthesis sample image, the prosthesis target includes at least one of a two-dimensional static image, a two-dimensional dynamic image, and a three-dimensional mold.

根據請求項6所述的方法，其中，在所述對至少兩幀所述待檢測圖像進行活體檢測，得到關於所述目標對象的活體檢測結果之後，所述方法還包括以下至少一步：在所述活體檢測結果為所述目標對象屬於活體的情況下，對其中一幀所述待檢測圖像進行目標識別，得到所述目標對象的識別結果；在所述活體檢測結果為所述目標對象不屬於活體的情況下，發送關於所述活體檢測結果的第一通知。The method according to claim 6, wherein after the living body detection is performed on the at least two frames of the to-be-detected images to obtain a living body detection result about the target object, the method further includes at least one of the following steps: In the case that the living body detection result is that the target object belongs to a living body, perform target recognition on one of the to-be-detected images to obtain a recognition result of the target object; In a case where the living body detection result is that the target object does not belong to a living body, a first notification about the living body detection result is sent.

根據請求項13所述的方法，其中，所述對其中一幀所述待檢測圖像進行目標識別，得到所述目標對象的識別結果，包括：對其中一幀所述待檢測圖像進行特徵提取，得到關於所述目標對象的目標特徵；獲取所述目標特徵分別與至少一個預存特徵的相似度；基於所述相似度，確定所述目標對象的識別結果。The method according to claim 13, wherein the performing target recognition on one of the to-be-detected images to obtain a recognition result of the target object includes: Perform feature extraction on one of the to-be-detected images to obtain target features about the target object; obtaining the similarity between the target feature and at least one pre-stored feature respectively; Based on the similarity, a recognition result of the target object is determined.

根據請求項13或14所述的方法，其中，在得到所述目標對象的識別結果之後，所述方法還包括：在所述識別結果為所述目標對象被識別成功的情況下，執行與所述目標對象的身份匹配的聯動控制；在所述識別結果為所述目標對象未被識別成功的情況下，發送關於所述識別結果的第二通知。The method according to claim 13 or 14, wherein after obtaining the identification result of the target object, the method further comprises: In the case that the recognition result is that the target object is successfully recognized, execute linkage control matching the identity of the target object; If the recognition result is that the target object is not successfully recognized, a second notification about the recognition result is sent.

根據請求項15所述的方法，其中，所述執行與所述目標對象的身份匹配的聯動控制，包括：在所述目標對象的身份屬於第一類身份，則控制關聯的門體打開，和/或，將所述目標對象的身份發送給關聯的第一通信設備，以使所述第一通信設備基於所述目標對象的身份進行與所述第一類身份相關的業務；在所述目標對象的身份屬於第二類身份，則控制外接設備發出警報，和/或，將所述目標對象的身份發送給所述第一通信設備，以使所述第一通信設備基於所述目標對象的身份進行與所述第二類身份相關的業務；所述發送關於所述活體檢測結果的第一通知，包括：將所述待檢測圖像和所述活體檢測結果中的至少一者進行第一編碼，並將所述第一編碼的結果打包至所述第一通知，以發送給第二通信設備；所述發送關於所述識別結果的第二通知，包括：將所述待檢測圖像和所述識別結果中的至少一者進行第二編碼，並將所述第二編碼的結果打包至所述第二通知，以發送給第三通信設備。The method according to claim 15, wherein the performing linkage control matching the identity of the target object includes: When the identity of the target object belongs to the first type of identity, the associated door is controlled to open, and/or the identity of the target object is sent to the associated first communication device, so that the first communication device is based on The identity of the target object performs services related to the first type of identity; When the identity of the target object belongs to the second type of identity, the external device is controlled to issue an alarm, and/or the identity of the target object is sent to the first communication device, so that the first communication device is based on the identity of the first communication device. The identity of the target object to carry out the business related to the second type of identity; The sending of the first notification about the living body detection result includes: performing first encoding on at least one of the to-be-detected image and the living body detection result, and packaging the result of the first encoding into the first notification for sending to the second communication device; The sending of the second notification about the identification result includes: Perform a second encoding on at least one of the image to be detected and the recognition result, and package the result of the second encoding into the second notification for sending to a third communication device.

一種電子設備，其中，包括記憶體和處理器，所述處理器配置為執行所述記憶體中儲存的程式指令，以實現請求項1至16任一項所述的方法。An electronic device, comprising a memory and a processor, the processor is configured to execute program instructions stored in the memory, so as to implement the method of any one of claim 1 to 16.

根據請求項17所述的設備，其中，所述電子設備為拍攝設備，所述拍攝設備還包括至少兩個圖像採集組件；所述處理器連接於所述圖像採集組件，以控制所述圖像採集組件的拍攝模式。The device according to claim 17, wherein the electronic device is a photographing device, and the photographing device further comprises at least two image acquisition components; the processor is connected to the image acquisition components to control the The capture mode of the image capture component.

根據請求項18所述的設備，其中，兩個所述圖像採集組件之間的基線距離為60毫米至150毫米。The device of claim 18, wherein the baseline distance between the two image capturing assemblies is 60 mm to 150 mm.

根據請求項18或19所述的設備，其中，所述拍攝設備還包括以下一者或多者：編解碼器，配置為對待發送給外部通信設備的第一資料進行編碼，以及對外部通信設備發送的第二資料進行解碼；聯動電路，配置為向外部關聯設備發送聯動控制指令；通信電路，配置為與外部通信設備進行通信。The device according to claim 18 or 19, wherein the photographing device further comprises one or more of the following: a codec, configured to encode the first data to be sent to the external communication device, and to decode the second data sent by the external communication device; A linkage circuit, configured to send linkage control instructions to external associated devices; A communication circuit configured to communicate with an external communication device.

一種電腦可讀儲存介質，其上儲存有程式指令，其中，所述程式指令被處理器執行時實現請求項1至16任一項所述的方法。A computer-readable storage medium on which program instructions are stored, wherein the program instructions implement the method described in any one of claim 1 to 16 when the program instructions are executed by a processor.