CN111767881A - Self-adaptive crowd density estimation device based on AI technology - Google Patents
Self-adaptive crowd density estimation device based on AI technology Download PDFInfo
- Publication number
- CN111767881A CN111767881A CN202010638776.1A CN202010638776A CN111767881A CN 111767881 A CN111767881 A CN 111767881A CN 202010638776 A CN202010638776 A CN 202010638776A CN 111767881 A CN111767881 A CN 111767881A
- Authority
- CN
- China
- Prior art keywords
- module
- density estimation
- crowd
- crowd density
- adaptive
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/44—Event detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The invention discloses an adaptive crowd density estimation device based on an AI (Artificial Intelligence) technology, which comprises a video data acquisition module, a target detection module, a crowd density estimation module, an adaptive module, an event early warning module, an event storage module and an event analysis module, wherein the video data acquisition module is used for acquiring real-time image pictures in a monitoring scene, the target detection module is used for detecting targets and positions in acquired video data and counting the crowd quantity of the targets, and the crowd density estimation module is used for analyzing and counting density maps by using the acquired real-time image data. This self-adaptation crowd density estimation device based on AI technique through adopting self-adaptation crowd density estimation device based on AI technique, no matter can all accurately calculate crowd density under crowd intensive or sparse scene, compares with current scheme, and robustness and generalization ability are better, and analysis speed is faster.
Description
Technical Field
The invention relates to the technical field of computer vision, in particular to an adaptive crowd density estimation device based on an AI (artificial intelligence) technology.
Background
In public environments such as stations, subways and airports, large-scale crowds are easy to appear, and in consideration of factors such as safety, flow dredging guidance is often needed at the moment to prevent safety accidents from happening, so that the method is particularly important for counting accurate crowd density in the scenes.
Reference is made to the patent "CN 201410339426" — a crowd density estimation method based on cascaded multi-stage convolutional neural network ", which is more relevant to the present invention:
the invention discloses a crowd density estimation method based on a cascade multi-level convolutional neural network, which comprises the following steps that 1) the multi-level convolutional neural network is adopted to extract characteristics from a low layer to a high layer, and the low layer characteristics and the high layer characteristics are combined together to form multi-stage characteristics, so that the separability of the crowd density characteristics is enhanced; 2) according to the similarity of the feature maps in the down-sampling layers of the multi-stage convolutional neural network, the connection of redundant neurons in the convolutional neural network is removed, so that the speed of feature extraction is increased; 3) according to the difficulty of the separability of the crowd density sample, two multilevel convolutional neural networks with different structures are trained, the two multilevel convolutional neural networks are cascaded in sequence from simple to complex to form a crowd density estimation model of the cascaded multilevel convolutional neural networks, and crowd density grades are quickly estimated on a detection image acquired by a video terminal in real time.
Reference is made to the patent "CN 201810567349" — a method for identifying and warning crowd density estimation "which is more relevant to the present invention:
the invention discloses a crowd density estimation identification and early warning method, and relates to the technical field of computer vision. The crowd density estimation identification and early warning method comprises the following steps: the collection of crowd's image and the detection of storage, crowd's head portrait density, the image contrast of crowd's density, the excessive value of crowd's density is reported to the police, the detection of crowd's sound density, the contrast of crowd's sound density and the excessive value of crowd's sound are reported to the police, camera device's output is connected with the input electricity of image collection module, image collection module and image storage module both way junction, the output of image collection module is connected with the input electricity of face identification module. According to the crowd density estimation identification and early warning method, the system can automatically remind each point position of abnormal conditions, and a deployment scheme can be made in advance by predicting the peak period of passenger flow.
In the prior art, many people observe the observation and then manage and control the observation, the method is time-consuming and labor-consuming, and quick and timely reaction cannot be made under the conditions of abnormity and the like; the target detection method is also used for carrying out crowd density statistics, but the method has poor effect under the conditions of small target, crowd and serious shielding; and the density map mode is used only, the effect is not ideal under the scene of sparse pedestrian flow, and the accuracy of the density estimation result is low.
Disclosure of Invention
Technical problem to be solved
Aiming at the defects of the prior art, the invention provides an adaptive crowd density estimation device based on an AI technology, which utilizes an advanced learning target detection method, a crowd density algorithm, and the like to analyze, model and adapt to a module, and overcomes the problem that the density estimation has larger errors when the crowd is sparse in the prior art.
(II) technical scheme
In order to achieve the purpose, the invention is realized by the following technical scheme: an adaptive crowd density estimation device based on an AI technology comprises a video data acquisition module, a target detection module, a crowd density estimation module, an adaptive module, an event early warning module, an event storage module and an event analysis module.
The video data acquisition module is used for acquiring real-time image pictures in a monitoring scene.
The target detection module is used for detecting targets and positions thereof in the collected video data and counting the number of people.
The crowd density estimation module utilizes the collected real-time image data to perform density map analysis statistics and calculate the crowd number.
The self-adaptive module is used for processing the mode of self-adaptive fusion of the target detection module and the crowd density estimation module
The event early warning module is used for outputting early warning to the analyzed abnormal events.
And the event storage module is used for storing the analyzed event video in time.
Preferably, the event analysis module is configured to further analyze the obtained crowd density data to determine whether there is an abnormal condition.
Preferably, after the target detection module receives the data from the video data acquisition module, the obtained image data is uniformly preprocessed, the processed image is processed by using a convolutional neural network to obtain a result, and the result is processed and analyzed to obtain the number of people.
Preferably, the crowd density estimation module receives real-time video data from the video data acquisition module, the preprocessing operation on the data is mainly mean value reduction, standard deviation and label removal processing is needed, the preprocessed data is subjected to feature extraction by utilizing a resnet50 network to obtain an output density map, the obtained density map is processed to obtain the crowd number, and in order to perform fusion in the self-adaptive module, the crowd density estimation module sends the density feature map obtained after network processing to the self-adaptive module for processing.
Preferably, the event early warning module outputs an early warning to the result analyzed by the self-adaptive module in time.
Preferably, the event storage module stores the analyzed result of the event early warning module to the system, and meanwhile, the analyzed result can be manually compared, and if the result is judged by mistake, the system automatically adds the analyzed result to the sample database to optimize the system model.
Preferably, in the target detection module, a feature map obtained after the convolutional neural network processing is used, and the feature map enters an adaptive module for further processing.
The invention also discloses a crowd density estimation algorithm system of the self-adaptive crowd density estimation device based on the AI technology, which specifically comprises the following steps:
s1, sending the obtained video data to a target detection module, a crowd density estimation module and a self-adaptive module by using a video data acquisition module;
s2, in the target detection module, for sparse people and with little pedestrian volume, using an anchor-free target method, passing the input image through a convolutional neural network to generate a heat map of key points, and predicting the height and width of a target boundary box by each peak value of the generated heat map;
s3, in the crowd density estimation module, when the pedestrian volume is dense, a crowd density estimation algorithm is used for generating a crowd density graph, and the crowd quantity is obtained by counting the density graph result;
s4, performing fusion analysis on the feature diagram result obtained by the target detection module and the feature result obtained by the crowd density estimation module through a self-adaptive module to obtain a final crowd density result;
and S5, processing and analyzing the analyzed crowd density result, if the analyzed result reaches an alarm threshold value, early warning is timely carried out through the event early warning module, and meanwhile, corresponding event video data are timely stored in the event storage module.
(III) advantageous effects
The invention provides an adaptive crowd density estimation device based on an AI (artificial intelligence) technology. Compared with the prior art, the method has the following beneficial effects:
(1) the adaptive crowd density estimation device based on the AI technology comprises a video data acquisition module, a target detection module, a crowd density estimation module, an adaptive module, an event early warning module, an event storage module and an event analysis module, can accurately calculate the crowd density no matter in a crowd dense or sparse scene by adopting the adaptive crowd density estimation device based on the AI technology, has high analysis speed, and utilizes the anchor-free deep learning target detection algorithm module and the crowd density estimation module to enable the reasoning time to reach real time.
(2) The adaptive crowd density estimation device based on the AI technology is high in accuracy and robustness, based on a deep learning algorithm and a large amount of collected sample data, and through a data augmentation technology, the system is high in generalization capability, robustness and detection accuracy.
(3) The adaptive crowd density estimation device based on the AI technology has the advantages that the system stability is high, the system runs in server resources, and a downtime prevention mechanism, a backup mechanism and a self-starting mechanism are arranged, so that the stable running of the system and the data safety are ensured.
Drawings
FIG. 1 is a block diagram of a system provided by an embodiment of the present invention;
FIG. 2 is a flowchart of a method of a target detection module and a crowd density estimation module according to an embodiment of the invention;
fig. 3 is a flowchart of an adaptive module method according to an embodiment of the present invention.
In the figure, a 101 video data acquisition module, a 102 target detection module, a 103 crowd density estimation module, a 104 self-adapting module, a 105 event early warning module and a 106 event storage module are adopted.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1-3, an embodiment of the present invention provides a technical solution: an adaptive crowd density estimation device based on an AI technology comprises a video data acquisition module 101, a target detection module 102, a crowd density estimation module 103, an adaptive module 104, an event early warning module 105, an event storage module 106 and an event analysis module.
The video data acquisition module 101 is configured to acquire real-time image frames in a monitoring scene.
The target detection module 102 is configured to detect targets and positions thereof in the acquired video data and count the number of people, after the target detection module 102 receives the data from the video data acquisition module 101, the target detection module 102 performs preprocessing on the acquired image data, then processes the processed image data by using a convolutional neural network to obtain a result, and analyzes the result to obtain the number of people, and in the target detection module 102, a feature map obtained after the convolutional neural network processing is used to enter the adaptive module 104 for further processing.
The crowd density estimation module 103 performs density map analysis statistics by using the acquired real-time image data to calculate the crowd number, the crowd density estimation module 103 receives the real-time video data from the video data acquisition module 101, the data needs to be preprocessed firstly, the preprocessing operation is mainly to reduce the mean value, the standard deviation and the label are removed, the preprocessed data is subjected to feature extraction by using a resnet50 network to obtain an output density map, the obtained density map is processed to obtain the crowd number, and in order to perform fusion in the adaptive module 104, the crowd density estimation module 103 sends the density feature map obtained after the network processing to the adaptive module 104 for processing.
In the existing density estimation method, good effects can be achieved in a dense scene, in an actual application scene, the dense scene is not always the case, the crowd quantity in the scene can be sparse in many times in different time periods, the traffic of people in the typical scenes such as a station and a ticket office can be large in holidays and holidays, and the traffic of people in the working days, early morning and other time periods can be sparse. In order to solve the problem in the actual scene, the adaptive module 104 is used in the invention, and the adaptive module 104 is used for processing the target detection module 102 and the crowd density estimation module 103 in an adaptive fusion mode, so that the crowd density result can be accurately estimated by the system no matter the crowd density is in a dense scene or in a sparse scene.
The event early warning module 105 is configured to output an early warning for the analyzed abnormal event, and the event early warning module 105 outputs an early warning in time for the result analyzed by the adaptive module 104.
The event storage module 106 is used for storing the analyzed event video in time, the event storage module 106 stores the analyzed result of the event early warning module 105 into the system according to the analyzed result reaching the early warning level, meanwhile, the analyzed result can be compared manually, if the result is judged by mistake, the system automatically adds the analyzed result into the sample database to optimize the system model, and therefore the generalization ability and robustness of the system are continuously updated, and the accuracy is improved.
In the invention, the event analysis module is used for further analyzing the obtained crowd density data and judging whether abnormal conditions exist, such as potential safety hazards caused by overlarge crowd density.
The invention also discloses a crowd density estimation algorithm system of the self-adaptive crowd density estimation device based on the AI technology, which specifically comprises the following steps:
s1, the video data acquisition module 101 is used for sending the obtained video data to the target detection module 102, the crowd density estimation module 103 and the self-adaptive module 104;
s2, in the target detection module 102, for sparse people with little pedestrian volume, the robustness of the crowd density algorithm directly used is poor, the prediction accuracy is poor, at the moment, the target detection method is more suitable for being used, the anchor-free target method is used, an input image is subjected to a convolutional neural network to generate a heat map of a key point, each peak value of the generated heat map predicts the height and width of a target boundary frame, the anchor-free target detection method for predicting a target by using a central point does not need to set the anchor and nms processes in advance, the configuration of hyper-parameters is greatly reduced, the accuracy is high, and real-time processing can be realized;
s3, in the crowd density estimation module 103, when the pedestrian volume is dense, a crowd density estimation algorithm is used for generating a crowd density graph, and the crowd number is obtained by counting the density graph result;
s4, performing fusion analysis on the feature diagram result obtained by the target detection module 102 and the feature result obtained by the crowd density estimation module 103 through the self-adaptive module 104 to obtain a final crowd density result;
and S5, processing and analyzing the analyzed crowd density result, if the analyzed result reaches an alarm threshold value, early warning is timely carried out through the event early warning module 105, and meanwhile, corresponding event video data are timely stored in the event storage module 106.
As shown in fig. 2, a method flow of the target detection module 102 and the crowd density estimation module 103 is specifically implemented as follows:
the module 201: the obtained video monitoring data are respectively input into the target detection module 102 and the crowd posture estimation module for processing;
a module 202, in which preprocessing operation is performed on input picture data, preprocessing flows in a target detection and crowd pose estimation module are basically consistent, normalization operation is mainly required to be performed on pictures, a Gaussian kernel is required to be used for generating a thermodynamic diagram of a label in a training stage, and in target detection, I ∈ R is enabledW*H*3For an input image, width W and height H, the goal is to generate a keypoint thermodynamic diagram
Where R is the output step size, default to use R-4, C is the number of classes,
which is indicative of the detected key point(s),
representing the background, and dispersing key points on the heat map through Gaussian kernels; in the density estimation, Gaussian processing is directly carried out on the key points of the human head, so that the key points are dispersed on a heat map;
the module 203: in the target detection module 102, the convolutional neural network uses features such as resnet, DLA and the like to extract a network, so as to obtain a feature map of a target key point; meanwhile, in the crowd density estimation module 103, feature extraction is also performed on input data by using a convolutional neural network to obtain a corresponding density feature map, and it should be noted that the output density map points need to be multiplied by an amplification factor, which is beneficial to model convergence;
the module 204: the module is mainly responsible for collecting the heat map finally obtained from the target detection module 102 and the density feature map obtained from the crowd density estimation module 103, and preparing the density feature map as the input of the adaptive module 104.
As shown in fig. 3, the method flow of the adaptive module 104 is specifically implemented as follows:
module 303: because the sizes of the feature maps and the number of channels from the two modules are different, the module mainly uses 1-by-1 convolution to adjust the number of the channels of the feature maps to be consistent, and then uses an interpolation mode to reduce the feature maps to the same size;
the module 304: in order to solve the problem that a pure density model in an actual scene is inaccurate in precision in a sparse scene and the situation that a single target is poor in detection effect in a dense scene, the module adaptively fuses target detection and density estimation and fuses the characteristics of the two; the specific treatment utilizes the formula:
y＝αFtarget detection+βFDensity estimation
Since F precedes block 304Target detectionTarget detection feature map and FDensity estimationThe density estimation profiles are unified to the same size and therefore can be added directly, for the weighting parameters α and β, the profiles after resize are convolved by 1 x 1, and the parameters α and β are all in the range of 0,1 by softmax after concat]And the sum is 1:
the module 305 performs statistics on the fused results to obtain a final crowd density estimate.
And those not described in detail in this specification are well within the skill of those in the art.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (8)
1. An adaptive crowd density estimation device based on AI technology is characterized in that: the system comprises a video data acquisition module (101), a target detection module (102), a crowd density estimation module (103), an adaptive module (104), an event early warning module (105), an event storage module (106) and an event analysis module;
the video data acquisition module (101) is used for acquiring real-time image pictures in a monitoring scene;
the target detection module (102) is used for detecting targets and positions thereof in the acquired video data and counting the number of people;
the crowd density estimation module (103) performs density map analysis statistics by using the collected real-time image data to calculate the crowd number;
the adaptive module (104) is used for processing the target detection module (102) and the crowd density estimation module (103) in an adaptive fusion mode;
the event early warning module (105) is used for outputting early warning to the analyzed abnormal events;
the event storage module (106) is used for storing the analyzed event video in time.
2. The AI-technology-based adaptive population density estimation device according to claim 1, wherein: and the event analysis module is used for further analyzing the obtained crowd density data and judging whether abnormal conditions exist or not.
3. The AI-technology-based adaptive population density estimation device according to claim 1, wherein: after the target detection module (102) receives the data from the video data acquisition module (101), the obtained image data is uniformly preprocessed, the processed image is processed by a convolutional neural network to obtain a result, and the result is processed and analyzed to obtain the number of people.
4. The AI-technology-based adaptive population density estimation device according to claim 1, wherein: the crowd density estimation module (103) receives real-time video data from the video data acquisition module (101), preprocessing operation is required to be performed on the data firstly, the preprocessing operation is mainly mean value reduction, standard deviation and label processing is removed, characteristics of the preprocessed data are extracted through a resnet50 network to obtain an output density map, the obtained density map is processed to obtain the crowd number, and in order to perform fusion in the self-adaptive module (104), the crowd density estimation module (103) uses the density characteristic map obtained after network processing to send the density characteristic map into the self-adaptive module (104) for processing.
5. The AI-technology-based adaptive population density estimation device according to claim 1, wherein: and the event early warning module (105) outputs an early warning in time for the result analyzed by the self-adapting module (104).
6. The AI-technology-based adaptive population density estimation device according to claim 1, wherein: the event storage module (106) stores the analyzed result of the event early warning module (105) into the system, meanwhile, the analyzed result can be compared manually, and if the result is judged by mistake, the system automatically adds the analyzed result into a sample database to optimize a system model.
7. The adaptive crowd density estimation device based on the AI technology according to claim 3, wherein: in the target detection module (102), a feature map obtained after the convolutional neural network processing is utilized, and the feature map enters an adaptive module (104) for further processing.
8. A crowd density estimation algorithm system of an AI technology based adaptive crowd density estimation apparatus according to any one of claims 1 to 7, wherein: the method specifically comprises the following steps:
s1, the video data acquisition module (101) is utilized to send the obtained video data to the target detection module (102), the crowd density estimation module (103) and the self-adapting module (104);
s2, in the target detection module (102), for sparse people and with little pedestrian volume, using an anchor-free target method, generating a heat map of key points after the input image passes through a convolutional neural network, and predicting the height and width of a target boundary box by each peak value of the generated heat map;
s3, in the crowd density estimation module (103), when the pedestrian volume is dense, a crowd density estimation algorithm is used for generating a crowd density graph, and the crowd number is obtained by counting the density graph result;
s4, performing fusion analysis on the feature diagram result obtained by the target detection module (102) and the feature result obtained by the crowd density estimation module (103) through the self-adaption module (104) to obtain a final crowd density result;
and S5, processing and analyzing the analyzed crowd density result, if the analyzed result reaches an alarm threshold value, early warning is timely carried out through the event early warning module (105), and meanwhile, corresponding event video data are timely stored in the event storage module (106).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010638776.1A CN111767881A (en) | 2020-07-06 | 2020-07-06 | Self-adaptive crowd density estimation device based on AI technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010638776.1A CN111767881A (en) | 2020-07-06 | 2020-07-06 | Self-adaptive crowd density estimation device based on AI technology |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111767881A true CN111767881A (en) | 2020-10-13 |
Family
ID=72724533
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010638776.1A Pending CN111767881A (en) | 2020-07-06 | 2020-07-06 | Self-adaptive crowd density estimation device based on AI technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111767881A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112580443A (en) * | 2020-12-02 | 2021-03-30 | 燕山大学 | Pedestrian detection method based on embedded device improved CenterNet |
| CN112783941A (en) * | 2021-01-07 | 2021-05-11 | 合肥工业大学 | Real-time detection method for large-scale group aggregation events |
| CN114627560A (en) * | 2022-05-13 | 2022-06-14 | 浙江大华技术股份有限公司 | Motion recognition method, motion recognition model training method and related device |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102136060A (en) * | 2011-03-03 | 2011-07-27 | 苏州市慧视通讯科技有限公司 | Method for detecting population density |
| US20150014526A1 (en) * | 2010-01-19 | 2015-01-15 | Rapiscan Systems, Inc. | Method of Electron Beam Transport in an X-Ray Scanner |
| CN107066963A (en) * | 2017-04-11 | 2017-08-18 | 福州大学 | A kind of adaptive people counting method |
| CN108717528A (en) * | 2018-05-15 | 2018-10-30 | 苏州平江历史街区保护整治有限责任公司 | A kind of global population analysis method of more strategies based on depth network |
| CN108960404A (en) * | 2017-05-22 | 2018-12-07 | 浙江宇视科技有限公司 | A kind of people counting method and equipment based on image |
| CN109657625A (en) * | 2018-12-23 | 2019-04-19 | 广东腾晟信息科技有限公司 | A kind of identification of crowd density estimation and method for early warning |
| CN110321869A (en) * | 2019-07-10 | 2019-10-11 | 应急管理部天津消防研究所 | Personnel's detection and extracting method based on Multiscale Fusion network |
| CN110618635A (en) * | 2019-10-08 | 2019-12-27 | 中兴飞流信息科技有限公司 | Train cab operation specification monitoring system based on AI technology |
| CN110942015A (en) * | 2019-11-22 | 2020-03-31 | 上海应用技术大学 | Crowd density estimation method |
| CN111160108A (en) * | 2019-12-06 | 2020-05-15 | 华侨大学 | Anchor-free face detection method and system |
-
2020
- 2020-07-06 CN CN202010638776.1A patent/CN111767881A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150014526A1 (en) * | 2010-01-19 | 2015-01-15 | Rapiscan Systems, Inc. | Method of Electron Beam Transport in an X-Ray Scanner |
| CN102136060A (en) * | 2011-03-03 | 2011-07-27 | 苏州市慧视通讯科技有限公司 | Method for detecting population density |
| CN107066963A (en) * | 2017-04-11 | 2017-08-18 | 福州大学 | A kind of adaptive people counting method |
| CN108960404A (en) * | 2017-05-22 | 2018-12-07 | 浙江宇视科技有限公司 | A kind of people counting method and equipment based on image |
| CN108717528A (en) * | 2018-05-15 | 2018-10-30 | 苏州平江历史街区保护整治有限责任公司 | A kind of global population analysis method of more strategies based on depth network |
| CN109657625A (en) * | 2018-12-23 | 2019-04-19 | 广东腾晟信息科技有限公司 | A kind of identification of crowd density estimation and method for early warning |
| CN110321869A (en) * | 2019-07-10 | 2019-10-11 | 应急管理部天津消防研究所 | Personnel's detection and extracting method based on Multiscale Fusion network |
| CN110618635A (en) * | 2019-10-08 | 2019-12-27 | 中兴飞流信息科技有限公司 | Train cab operation specification monitoring system based on AI technology |
| CN110942015A (en) * | 2019-11-22 | 2020-03-31 | 上海应用技术大学 | Crowd density estimation method |
| CN111160108A (en) * | 2019-12-06 | 2020-05-15 | 华侨大学 | Anchor-free face detection method and system |
Non-Patent Citations (4)
| Title |
|---|
| JIANG LIU等: "DecideNet: Counting Varying Density Crowds Through Attention Guided Detection and Density Estimation", 《CVPR 2018》, pages 5197 - 5206 * |
| MINGLIANG XU等: "Depth Information Guided Crowd Counting for Complex Crowd Scenes", 《ARXIV:1803.02256V2》, pages 1 - 9 * |
| 徐扬: "基于深度学习的行人遮挡检测研究", 《中国优秀硕士学位论文全文数据库 (信息科技辑)》, no. 2020, pages 138 - 148 * |
| 王新雨等: "基于深度学习的密集人群安全监测***", 《物联网技术 》, vol. 9, no. 11, pages 8 - 12 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112580443A (en) * | 2020-12-02 | 2021-03-30 | 燕山大学 | Pedestrian detection method based on embedded device improved CenterNet |
| CN112580443B (en) * | 2020-12-02 | 2022-03-15 | 燕山大学 | Pedestrian detection method based on embedded device improved CenterNet |
| CN112783941A (en) * | 2021-01-07 | 2021-05-11 | 合肥工业大学 | Real-time detection method for large-scale group aggregation events |
| CN114627560A (en) * | 2022-05-13 | 2022-06-14 | 浙江大华技术股份有限公司 | Motion recognition method, motion recognition model training method and related device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110428522B (en) | Intelligent security system of wisdom new town | |
| Aboah | A vision-based system for traffic anomaly detection using deep learning and decision trees | |
| US7639840B2 (en) | Method and apparatus for improved video surveillance through classification of detected objects | |
| CN111767881A (en) | Self-adaptive crowd density estimation device based on AI technology | |
| CN110738127A (en) | Helmet identification method based on unsupervised deep learning neural network algorithm | |
| Patil et al. | Fggan: A cascaded unpaired learning for background estimation and foreground segmentation | |
| CN111402298A (en) | Grain depot video data compression method based on target detection and trajectory analysis | |
| KR102391853B1 (en) | System and Method for Processing Image Informaion | |
| CN113963301A (en) | Space-time feature fused video fire and smoke detection method and system | |
| CN112132048A (en) | Community patrol analysis method and system based on computer vision | |
| CN115691034A (en) | Intelligent household abnormal condition warning method, system and storage medium | |
| CN114267082A (en) | Bridge side falling behavior identification method based on deep understanding | |
| Vijayan et al. | A fully residual convolutional neural network for background subtraction | |
| CN113920585A (en) | Behavior recognition method and device, equipment and storage medium | |
| CN117197726B (en) | Important personnel accurate management and control system and method | |
| CN111160150A (en) | Video monitoring crowd behavior identification method based on depth residual error neural network convolution | |
| CN111639585A (en) | Self-adaptive crowd counting system and self-adaptive crowd counting method | |
| CN108960181B (en) | Black smoke vehicle detection method based on multi-scale block LBP and hidden Markov model | |
| CN115294519A (en) | Abnormal event detection and early warning method based on lightweight network | |
| CN113822155A (en) | Clustering-assisted weak surveillance video anomaly detection method and device | |
| CN106874885A (en) | A kind of crowd's method for detecting abnormality based on energy level changes in distribution | |
| CN113627383A (en) | Pedestrian loitering re-identification method for panoramic intelligent security | |
| CN113255549A (en) | Intelligent recognition method and system for pennisseum hunting behavior state | |
| CN112241694A (en) | Method for identifying unworn safety belt based on CenterNet | |
| KR20210114169A (en) | Method for analyzing monitoring image using object verification, and apparatus for the same |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |