WO2018121006A1 - 一种车牌定位方法及装置 - Google Patents

一种车牌定位方法及装置 Download PDF

Info

Publication number
WO2018121006A1
WO2018121006A1 PCT/CN2017/106661 CN2017106661W WO2018121006A1 WO 2018121006 A1 WO2018121006 A1 WO 2018121006A1 CN 2017106661 W CN2017106661 W CN 2017106661W WO 2018121006 A1 WO2018121006 A1 WO 2018121006A1
Authority
WO
WIPO (PCT)
Prior art keywords
license plate
area
image
plate image
combination
Prior art date
Application number
PCT/CN2017/106661
Other languages
English (en)
French (fr)
Inventor
蔡晓蕙
Original Assignee
杭州海康威视数字技术股份有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 杭州海康威视数字技术股份有限公司 filed Critical 杭州海康威视数字技术股份有限公司
Priority to EP17889416.8A priority Critical patent/EP3564855A4/en
Priority to US16/473,787 priority patent/US11126882B2/en
Publication of WO2018121006A1 publication Critical patent/WO2018121006A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
    • G06V20/63—Scene text, e.g. street names
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/14—Image acquisition
    • G06V30/148—Segmentation of character regions
    • G06V30/153—Segmentation of character regions using recognition of characters or words
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
    • G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
    • G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
    • G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
    • G06V20/625—License plates

Definitions

  • the present application relates to the field of intelligent transportation technologies, and in particular, to a license plate positioning method and device.
  • the license plate is the “identity card” of the vehicle and is an important information that distinguishes it from other motor vehicles.
  • the license plate recognition technology has been widely used in scenes such as bayonet, parking lot and electronic police to obtain the license plate information of vehicles in the scene, and exerts the power of “intelligent traffic algorithm” in many aspects such as public security management.
  • the license plate image of the license plate image including the vehicle number plate when used, the color of the character and the background in the license plate is different, and the license plate area has a plurality of features such as a boundary point of the character and the background to be positioned.
  • the horizontal coordinate of the pixel point when positioning the license plate area therein, the horizontal coordinate of the pixel point may be the horizontal axis, and the pixel value of the pixel point may be the vertical axis, and the pixels of each row of the license plate image may be obtained.
  • the value change curve finds a portion of the pixel value regular fluctuation change from the pixel value change curve, and the area on the corresponding license plate image is the license plate area.
  • the license plate location method described above can locate the license plate area in a common license plate image.
  • a license plate with a small number of characters there is little boundary between the character and the background in the license plate area.
  • the number of consecutive license plate characters in the license plate shown in Fig. 1b is only one or two.
  • the portion where the pixel value regularly changes is not obvious, so the accuracy of the license plate area positioned by the above license plate positioning method is not high.
  • the purpose of the embodiment of the present application is to provide a license plate positioning method and device to improve the accuracy of the license plate location process.
  • the specific technical solution is as follows.
  • the embodiment of the present application discloses a license plate location method, and the method includes:
  • the target network includes a feature extraction layer and a regression layer, the target network is trained in advance by a sample license plate image, and the sample license plate image includes positive a sample license plate image, the positive sample license plate image including a combination of a license plate frame area and a license plate character area;
  • the determining, according to the positioning area corresponding to the obtained combination, the license plate area of the license plate image to be located including:
  • the determining, according to the target positioning area, the license plate area of the license plate image to be located including:
  • the target positioning area in which the relative position is within a preset relative position range is determined as the license plate area of the license plate image to be positioned.
  • the determining, according to the target positioning area, the license plate area of the license plate image to be located including:
  • the license plate character area in the target positioning area is determined as the license plate area of the license plate image to be positioned.
  • the method further includes:
  • Character recognition is performed on the determined license plate area, and the license plate number of the license plate image to be positioned is obtained.
  • the embodiment of the present application discloses a license plate locating device, and the device includes:
  • An image obtaining module configured to obtain a license plate image to be located
  • An image sending module configured to send the to-be-positioned license plate image to a target network for license plate location, wherein the target network includes a feature extraction layer and a regression layer, and the target network is trained in advance by using a sample license plate image.
  • the sample license plate image includes a positive sample license plate image, and the positive sample license plate image includes a combination of a license plate frame area and a license plate character area;
  • a feature extraction module configured to extract, by the feature extraction layer, a feature value of the license plate image to be located, and send the feature value to the regression layer;
  • a combination obtaining module configured to obtain a combination of a license plate frame area and a license plate character area obtained by the regression layer according to the feature value
  • the area determining module is configured to determine a license plate area of the license plate image to be located according to the obtained positioning area corresponding to the combination.
  • the area determining module includes:
  • a target area determining sub-module configured to determine, as the target positioning area, a positioning area in which the obtained confidence in the combination is greater than a preset threshold
  • the license plate area determining submodule is configured to determine a license plate area of the license plate image to be located according to the target positioning area.
  • the license plate area determining submodule is specifically configured to:
  • the target positioning area in which the relative position is within a preset relative position range is determined as the license plate area of the license plate image to be positioned.
  • the license plate area determining sub-module is specifically configured to: determine a license plate character area in the target positioning area as a license plate area of the license plate image to be located.
  • the device further includes:
  • a license plate recognition module configured to perform character recognition on the determined license plate area, to obtain the to-be-positioned The license plate number of the license plate image.
  • the embodiment of the present application discloses an electronic device, which is suitable for vehicle license plate location, and the electronic device includes:
  • the circuit board is disposed inside the space enclosed by the housing, the processor and the memory are disposed on the circuit board; and the power supply circuit is used for each circuit of the electronic device or The device is powered;
  • the memory is for storing executable program code;
  • the processor runs the program corresponding to the executable program code by reading the executable program code stored in the memory for performing the following steps:
  • the target network includes a feature extraction layer and a regression layer, the target network is trained in advance by a sample license plate image, and the sample license plate image includes positive a sample license plate image, the positive sample license plate image including a combination of a license plate frame area and a license plate character area;
  • the processor determines, as the target positioning area, a positioning area in which the obtained confidence in the combination is greater than a preset threshold; and determines a license plate area of the license plate image to be located according to the target positioning area.
  • the processor determines a relative position between the license plate frame area and the license plate character area in the target positioning area; and the target positioning area in which the relative position is within a preset relative position range is determined to be The license plate area where the license plate image is located.
  • the embodiment of the present application discloses an application program, which is used to execute the license plate location method provided by the embodiment of the present application at runtime.
  • the License plate location methods include:
  • the target network includes a feature extraction layer and a regression layer, the target network is trained in advance by a sample license plate image, and the sample license plate image includes positive a sample license plate image, the positive sample license plate image including a combination of a license plate frame area and a license plate character area;
  • the embodiment of the present application discloses a storage medium for storing executable code, which is used to execute the license plate location method provided by the embodiment of the present application.
  • the license plate location method includes:
  • the target network includes a feature extraction layer and a regression layer, the target network is trained in advance by a sample license plate image, and the sample license plate image includes positive a sample license plate image, the positive sample license plate image including a combination of a license plate frame area and a license plate character area;
  • the license plate image to be located is sent to the target network for license plate location, wherein the target network includes a feature extraction layer and a regression layer; Then, the feature extraction layer extracts the feature value of the license plate image to be located, and sends the feature value to the regression layer, and obtains a combination of the license plate frame region and the license plate character region obtained by the regression layer according to the feature value; finally, according to the obtained combination a positioning area for determining a license plate area of the license plate image to be positioned.
  • the target network is trained in advance by the sample license plate image
  • the sample license plate image includes a positive sample license plate image
  • the positive sample license plate image includes a combination of the license plate frame area and the license plate character area.
  • the embodiment of the present application can determine the license plate area of the license plate image to be located according to the positioning area corresponding to the combination of the license plate frame area and the license plate character area detected by the target network. Since there may be many non-license card character areas in the license plate image to be located, according to the combination relationship between the license plate frame area and the license plate character area, the license plate area can be accurately located from many interference factors, and there is no need to distinguish the characters from the background. The positioning is performed, so that the solution provided by the embodiment of the present application can improve the accuracy of the license plate location process.
  • Figure 1a is an example of a license plate image including a license plate area
  • Figure 1b is an example of a license plate with a small number of characters
  • Figure 1c is an example of a license plate image of a vehicle body containing non-license characters
  • FIG. 2 is a schematic flow chart of a license plate location method according to an embodiment of the present application.
  • 3 is an example of a combination of a license plate frame area and a license plate character area in several license plates;
  • FIG. 4 is a schematic structural diagram of a target network
  • FIG. 5 is a schematic structural diagram of a license plate locating device according to an embodiment of the present application.
  • the embodiment of the present invention provides a method and a device for locating a license plate, which are applied to an electronic device, and the electronic device may be a terminal device or a server.
  • the terminal device may include a computer, a tablet computer, a smart phone, a driving recorder, and the like. Applying the technical solution in the embodiment of the present application to perform license plate location can improve the accuracy of the license plate location process.
  • the present application will be described in detail below through specific embodiments.
  • FIG. 2 is a schematic flowchart of a license plate location method according to an embodiment of the present disclosure, which is applied to an electronic device. The method specifically includes the following steps:
  • Step S201 Obtain a license plate image to be located.
  • step S201 is a step performed by the execution subject.
  • the electronic device as the execution subject may include an image capturing device or may not include an image capturing device.
  • the electronic device may include: receiving the image of the license plate to be positioned collected by the image capturing device when obtaining the image of the license plate to be positioned.
  • the electronic device When the electronic device as the execution subject does not include the image capturing device, the electronic device may be connected to the external image capturing device.
  • the electronic device may include: acquiring the image of the license plate to be positioned collected by the image collecting device. .
  • the acquired license plate image to be located may be collected by the image acquisition device in real time, or may not be collected in real time, but stored by the image acquisition device after being collected in advance.
  • the above-mentioned license plate image to be positioned can be understood as an image to be positioned in the license plate area. It can be understood that the license plate is usually installed or placed on the vehicle. Therefore, the above-mentioned license plate image to be positioned can be understood as an image containing the vehicle to be positioned in the license plate area. Based on this, the image of the license plate to be positioned may be an image containing a vehicle captured on the road, and may be a car included in the parking lot. The image of the car, and so on. Certainly, the image of the license plate to be positioned may be any image including a license plate area, and the manner of obtaining the image of the license plate to be positioned in the present application is not specifically limited.
  • Step S202 Send the to-be-positioned license plate image to a target network for license plate location.
  • step S202 is a step performed by the execution subject.
  • the target network includes a feature extraction layer and a regression layer.
  • the feature extraction layer is configured to extract feature values of the license plate image to be located, and input the feature values into the regression layer.
  • the regression layer is configured to obtain a combination of a license plate frame area and a license plate character area according to the feature value.
  • the target network is pre-trained by a sample license plate image comprising a positive sample license plate image, the positive sample license plate image comprising a combination of a license plate frame area and a license plate character area.
  • the above-mentioned license plate frame area can be understood as an area enclosed by the outer frame of the license plate.
  • the license plate character area can be understood as the area where the license plate characters are located in the license plate area.
  • sending the to-be-positioned license plate image to the target network for license plate location may include: transmitting the to-be-positioned license plate image to a feature extraction layer in a target network for license plate location.
  • the target network may be a deep learning network such as a convolutional neural network.
  • the target network may further include an input layer, the input layer may be configured to receive the license plate image to be located, preprocess the license plate image to be positioned, and input the preprocessed license plate image to be input into the feature extraction layer.
  • the pre-processing may specifically include at least one of: normalizing pixel values of an image; processing grayscale values of image pixels according to a preset gray threshold to improve image over-bright or too dark The situation; modify the size of the image, etc.
  • transmitting the to-be-positioned license plate image to the target network for license plate location may include: transmitting the to-be-positioned license plate image to an input layer in a target network for license plate location.
  • the target network may further include an output layer.
  • the regression layer obtains a combination of the license plate frame area and the license plate character area
  • the combination can be output through the output layer.
  • the output layer may perform certain processing on the above combination and output it, such as data encapsulation.
  • the foregoing target network may include an input layer, a feature extraction layer, a regression layer, an output layer, and the like.
  • the target network is also It can be called a concatenated convolutional neural network.
  • the convolutional kernel and activation function in concatenated convolutional neural networks it can extract more complex character features, increase the inter-class difference between character features and non-character features, and also reduce character features and non-characters.
  • the intraclass difference between features makes it easier to return the regression layer to the license plate character area.
  • FIG. 4 shows a structural diagram of a target network, where the target network includes: an input layer 1, a feature extraction layer 2, a regression layer 3, and an output layer 4.
  • the license plate positioning is performed, the license plate image to be positioned is input to the input layer 1, and a combination of the license plate frame area and the license plate character area for the license plate image to be positioned is obtained from the output layer 4, and the combination is the output result of the output layer.
  • the output result may include a positioning area corresponding to the combination of the license plate frame area and the license plate character area and a corresponding confidence degree.
  • the output may contain one or more location areas.
  • the target network is pre-trained.
  • the sample license plate image may also include a negative sample license plate image when training the model, and the negative sample license plate image may be understood as an image that does not include the license plate area, that is, does not include the license plate frame area. And an image of the license plate character area.
  • a plurality of sample license plate images including a positive sample license plate image and a negative sample license plate image may be obtained in advance, wherein the positive sample license plate image carries a combination mark of the license plate frame area and the license plate character area; and the sample license plate is sent.
  • Image to feature extraction layer After receiving the sample license plate image, the feature extraction layer obtains the feature value of the sample license plate image, and transmits the feature value and the sample license plate image to the regression layer.
  • the regression layer receives the information sent by the feature extraction layer, and determines a suspected location area from the sample license plate image according to the information, and then calculates a coincidence degree between the suspected location area and the standard location area, where the standard location area is the marked license plate frame area.
  • a positioning area corresponding to the combination of the license plate character area is continuously adjusted according to the comparison result.
  • the target network training can be considered completed.
  • the main field area may be selected as the license plate character area for marking.
  • the above main field can be understood as a character part having the following characteristics: the character size is large and distributed in the middle area of the license plate.
  • the above-mentioned slave field can be understood as a character part having the following characteristics: the character size is smaller than the main field character, and is distributed in the car. At the edge of the card, it has the meaning of identification (identifying the characteristics of the country or region).
  • the first license plate includes two characters "5", the left character “5" indicates the region to which the vehicle belongs, and is the slave field; the right character “5" indicates the license plate.
  • the number is the main field.
  • the character “36" is the main field and the character “N” is the slave field.
  • Step S203 The feature extraction layer extracts a feature value of the license plate image to be located, and sends the feature value to the regression layer.
  • step S203 is a step performed by the feature extraction layer.
  • obtaining the feature value of the to-be-positioned license plate image extracted by the feature extraction layer may include: obtaining a feature value extracted by the feature extraction layer for the pixel in the to-be-positioned license plate image.
  • step S203 may include the following implementation manners:
  • the image of the license plate to be positioned after the input layer is preprocessed is obtained, and the pre-processed license plate image to be located is sent to the feature extraction layer, and the feature of the pre-processed license plate image extracted by the feature extraction layer is obtained. Value and send the feature value to the regression layer.
  • the input layer preprocesses the license plate image to be positioned, and transmits the preprocessed license plate image to be located to the feature extraction layer.
  • the license plate image to be located sent by the input layer is obtained, and the license plate image to be located is sent to the feature extraction layer, the feature value of the license plate image to be located extracted by the feature extraction layer is acquired, and the feature value is sent to the regression.
  • the input layer does not pre-process the license plate image, but directly sends the license plate image to be located to the feature extraction layer.
  • Step S204 Obtain a combination of the license plate frame area and the license plate character area obtained by the regression layer according to the feature value.
  • step S204 is a step performed by the execution subject.
  • the regression layer also obtains the confidence level corresponding to each combination when the combination of the license plate frame area and the license plate character area is obtained according to the feature value. Therefore, the confidence of each combination can also be included in the combination obtained above.
  • step S204 Can include:
  • the obtained combination is sent to the output layer such that the output layer outputs the obtained combination.
  • the method may further include: processing the obtained combination according to a preset rule, and outputting the processed combination.
  • the preset rules may include rules such as sorting according to a preset format.
  • Step S205 Determine a license plate area of the license plate image to be located according to the obtained positioning area corresponding to the combination.
  • step S205 is a step performed by the execution body, and step S205 may include: determining, as the target location area, a location area in which the obtained confidence in the combination is greater than a preset threshold, and determining the to-be-determined according to the target location area.
  • the license plate area of the license plate image is a step performed by the execution body, and step S205 may include: determining, as the target location area, a location area in which the obtained confidence in the combination is greater than a preset threshold, and determining the to-be-determined according to the target location area.
  • the obtained combination includes combination 1, combination 2, combination 3, and combination 4, each combination being a combination of a license plate frame area and a license plate character area, and the confidence of the four combinations
  • the degrees are 320, 270, 950 and 890, respectively.
  • the preset threshold is 800, and it is determined that the confidence levels of the combination 3 and the combination 4 are both greater than a preset threshold, and the positioning areas corresponding to the combination 3 and the combination 4 can be determined as the target positioning area.
  • determining the license plate area of the license plate image to be located according to the target positioning area may include the following implementation manners:
  • a relative position between the license plate frame area and the license plate character area in the target positioning area is determined, and the target positioning area in which the relative position is within a preset relative position range is determined as the license plate image to be located. License plate area.
  • the relative position between the license plate frame area and the license plate character area may be represented by the distance between the corresponding boundaries of the two regions, or the relative positions of the center points of the two regions may be used, and the length ratio and width of the two regions may be combined. The ratio is expressed.
  • the relative position range may be determined according to the relative position between the license plate frame area and the license plate character area in the pre-acquired positive sample license plate image. It can be understood that the output result is filtered according to the preset relative position range, and the license plate area can be effectively filtered from the obtained combination, and the accuracy of the license plate positioning process is improved.
  • the license plate character area in the target positioning area is determined as the license plate area of the license plate image to be located.
  • the license plate frame area contains more image information than the license plate character area, and also contains a lot of interference information except the license plate characters. Therefore, in order to improve the accuracy of the license plate location process, the license plate character area in the target positioning area may be directly determined as the license plate area of the license plate image to be located.
  • the slave field portion of the license plate may be further determined for the license plate frame area in the target positioning area.
  • the license plate category can be determined by the branch of the regression layer, that is, the license plate category is determined according to the overall texture information in the license plate frame, and then the positional relationship between the field and the main field is determined by the preset license plate category, thereby determining from the license plate frame region. The part of the license plate from the field.
  • the license plate image to be located is sent to the target network for license plate location, wherein the target network includes a feature extraction layer and a regression layer;
  • the feature extraction layer extracts the feature value of the license plate image to be located, and sends the feature value to the regression layer, and obtains a combination of the license plate frame region and the license plate character region obtained by the regression layer according to the feature value; finally, according to the obtained positioning corresponding to the combination
  • the area determines a license plate area of the license plate image to be located.
  • the target network is trained in advance by the sample license plate image, the sample license plate image includes a positive sample license plate image, and the positive sample license plate image includes a combination of the license plate frame area and the license plate character area.
  • the license plate area of the license plate image to be located is determined according to the positioning area corresponding to the combination of the license plate frame area and the license plate character area detected by the target network. Since there may be many non-license card character areas or non-license card frame areas in the license plate image to be located, for example, in the license plate image shown in FIG. 1c, there are some other text parts and area frame parts other than the vehicle license plate area; According to the combination relationship between the license plate frame area and the license plate character area, the license plate area can be accurately located from a plurality of interference factors, and the positioning is not required according to the boundary point between the character and the background. Therefore, the solution of the embodiment can improve the license plate location. The accuracy of the process.
  • the method may further include performing character recognition on the determined license plate area to obtain the license plate to be positioned.
  • the license plate number of the image may be performed by the method.
  • FIG. 5 is a schematic structural diagram of a license plate locating device according to an embodiment of the present disclosure.
  • the device is applied to an electronic device corresponding to the method embodiment shown in FIG. 2, and the device includes:
  • An image obtaining module 501 configured to obtain a license plate image to be located
  • the image sending module 502 is configured to send the to-be-positioned license plate image to a target network for license plate location, where the target network includes a feature extraction layer and a regression layer, and the target network is trained in advance by using a sample license plate image.
  • the sample license plate image includes a positive sample license plate image, and the positive sample license plate image includes a combination of a license plate frame area and a license plate character area;
  • a feature extraction module 503 configured to extract, by the feature extraction layer, a feature value of the license plate image to be located, and send the feature value to the regression layer;
  • a combination obtaining module 504 configured to obtain a combination of a license plate frame area and a license plate character area obtained by the regression layer according to the feature value;
  • the area determining module 505 is configured to determine a license plate area of the license plate image to be located according to the obtained positioning area corresponding to the combination.
  • the area determining module 505 may include:
  • a target area determining sub-module (not shown in the figure), configured to determine, as the target positioning area, a positioning area in which the obtained confidence in the combination is greater than a preset threshold;
  • the license plate area determining sub-module (not shown) is configured to determine a license plate area of the license plate image to be located according to the target positioning area.
  • the license plate area determining sub-module may be specifically used to:
  • the license plate area determining sub-module is specifically configured to:
  • the license plate character area in the target positioning area is determined as the license plate area of the license plate image to be positioned.
  • the device may further include: a license plate recognition module (not shown) for performing character recognition on the determined license plate area, and obtaining the to-be-positioned license plate The license plate number of the image.
  • a license plate recognition module (not shown) for performing character recognition on the determined license plate area, and obtaining the to-be-positioned license plate The license plate number of the image.
  • the description is relatively simple, and the relevant parts can be referred to the description of the method embodiment.
  • the embodiment of the present application provides an electronic device, which is suitable for license plate location, and the electronic device includes:
  • the circuit board is disposed inside the space enclosed by the housing, the processor and the memory are disposed on the circuit board; and the power supply circuit is used for each circuit of the electronic device or The device is powered;
  • the memory is for storing executable program code;
  • the processor runs the program corresponding to the executable program code by reading the executable program code stored in the memory for performing the following steps:
  • the target network includes a feature extraction layer and a regression layer, the target network is trained in advance by a sample license plate image, and the sample license plate image includes positive a sample license plate image, the positive sample license plate image including a combination of a license plate frame area and a license plate character area;
  • the electronic device can exist in various forms, including but not limited to:
  • Mobile communication devices These devices are characterized by mobile communication functions and are mainly aimed at providing voice and data communication.
  • Such terminals include: smart phones (such as iPhone), multimedia phones, functional phones, and low-end phones.
  • Ultra-mobile personal computer equipment This type of equipment belongs to the category of personal computers, has computing and processing functions, and generally has mobile Internet access.
  • Such terminals include: PDAs, MIDs, and UMPC devices, such as the iPad.
  • Portable entertainment devices These devices can display and play multimedia content. Such devices include: audio, video players (such as iPod), handheld game consoles, e-books, and smart toys and portable car navigation devices.
  • the server consists of a processor, a hard disk, a memory, a system bus, etc.
  • the server is similar to a general-purpose computer architecture, but because of the need to provide highly reliable services, processing power and stability High reliability in terms of reliability, security, scalability, and manageability.
  • the license plate area of the license plate image to be located can be determined according to the positioning area corresponding to the combination of the license plate frame area and the license plate character area detected by the target network. Since there may be many non-license card character areas in the license plate image to be located, according to the combination relationship between the license plate frame area and the license plate character area, the license plate area can be accurately located from many interference factors, and there is no need to distinguish the characters from the background. The positioning is performed, so that the solution provided by the embodiment can be used to improve the accuracy of the license plate location process.
  • the processor determines, as the target location area, a location area in which the obtained confidence in the combination is greater than a preset threshold; and determines the to-be-positioned according to the target location area.
  • the license plate area of the license plate image is determined, as the target location area, a location area in which the obtained confidence in the combination is greater than a preset threshold; and determines the to-be-positioned according to the target location area.
  • the processor determines a relative position between the license plate frame area and the license plate character area in the target positioning area; and the relative position is within a preset relative position range.
  • the target location area is determined as the license plate area of the license plate image to be located.
  • the embodiment of the present application provides an application program for executing the license plate location method provided by the embodiment of the present application at runtime.
  • the license plate location method includes:
  • the target network includes a feature extraction layer and a regression layer, the target network is trained in advance by a sample license plate image, and the sample license plate image includes positive a sample license plate image, the positive sample license plate image including a combination of a license plate frame area and a license plate character area;
  • the license plate area of the license plate image to be located can be determined according to the positioning area corresponding to the combination of the license plate frame area and the license plate character area detected by the target network. Since there may be many non-license card character areas in the license plate image to be located, according to the combination relationship between the license plate frame area and the license plate character area, the license plate area can be accurately located from many interference factors, and there is no need to distinguish the characters from the background. The positioning is performed, so that the solution provided by the embodiment can be used to improve the accuracy of the license plate location process.
  • the embodiment of the present application provides a storage medium for storing executable code, which is used to execute the license plate location method provided by the embodiment of the present application.
  • the license plate location method includes:
  • the target network includes a feature extraction layer and a regression layer, the target network is trained in advance by a sample license plate image, and the sample license plate image includes positive a sample license plate image, the positive sample license plate image including a combination of a license plate frame area and a license plate character area;
  • the license plate area of the license plate image to be located can be determined according to the positioning area corresponding to the combination of the license plate frame area and the license plate character area detected by the target network. Since there may be many non-license card character areas in the license plate image to be located, according to the combination relationship between the license plate frame area and the license plate character area, the license plate area can be accurately located from many interference factors, and there is no need to distinguish the characters from the background. The positioning is performed, so that the solution provided by the embodiment can be used to improve the accuracy of the license plate location process.
  • the storage medium referred to herein means a ROM/RAM, a magnetic disk, an optical disk, or the like.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • Probability & Statistics with Applications (AREA)
  • Algebra (AREA)
  • Operations Research (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Traffic Control Systems (AREA)
  • Image Analysis (AREA)
  • Character Input (AREA)

Abstract

本申请实施例提供了一种车牌定位方法及装置。所述方法包括：获得待定位车牌图像；发送所述待定位车牌图像至用于车牌定位的目标网络，所述目标网络包括特征提取层和回归层；所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。其中，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合。应用本申请实施例提供的方案进行车牌定位，能够提高车牌定位过程的准确性。

Description

一种车牌定位方法及装置
本申请要求于2016年12月30日提交中国专利局、申请号为201611261487.4、发明名称为“一种车牌定位方法及装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及智能交通技术领域，特别涉及一种车牌定位方法及装置。
背景技术
车牌是车辆的“身份证”，是区别于其他机动车辆的一项重要信息。车牌识别技术已被广泛应用在卡口、停车场和电子警察等场景中，以获取场景内车辆的号牌信息，在治安管理等众多方面发挥着“智能交通算法”的威力。
相关技术中，在对包含车辆号牌的车牌图像进行车牌定位时，通常根据车牌中字符与背景的颜色不同，车牌区域存在大量字符与背景的交界点等特征进行定位。例如，对于图1a所示的车牌图像，在定位其中的车牌区域时，可以以像素点的横向坐标为横轴，以像素点的像素值为纵轴，获得车牌图像的每一行像素点的像素值变化曲线，从该像素值变化曲线中找到像素值规律性起伏变化的部分，该部分对应的车牌图像上的区域即为车牌区域。
通常，采用上述车牌定位方法能够定位出常见车牌图像中的车牌区域。但是，对于字符数量很少的车牌，车牌区域中字符与背景的交界点很少，例如图1b中所示的车牌中连续车牌字符数量只有1～2个。这类车牌对应的车牌图像的像素值变化曲线中，像素值规律性起伏变化的部分不明显，因此采用上述车牌定位方法定位的车牌区域的准确性不高。
发明内容
本申请实施例的目的在于提供了一种车牌定位方法及装置，以提高车牌定位过程的准确性。具体的技术方案如下。
为了达到上述目的，一方面，本申请实施例公开了一种车牌定位方法，所述方法包括：
获得待定位车牌图像；
发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
可选的，所述根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域，包括：
将所获得的组合中置信度大于预设阈值的定位区域确定为目标定位区域；
根据所述目标定位区域，确定所述待定位车牌图像的车牌区域。
可选的，所述根据所述目标定位区域，确定所述待定位车牌图像的车牌区域，包括：
确定所述目标定位区域中车牌框区域和车牌字符区域之间的相对位置；
将所述相对位置处于预设的相对位置范围内的目标定位区域，确定为所述待定位车牌图像的车牌区域。
可选的，所述根据所述目标定位区域，确定所述待定位车牌图像的车牌区域，包括：
将所述目标定位区域中的车牌字符区域，确定为所述待定位车牌图像的车牌区域。
可选的，所述方法还包括：
对确定的车牌区域进行字符识别，获得所述待定位车牌图像的车牌号码。
为了达到上述目的，另一方面，本申请实施例公开了一种车牌定位装置，所述装置包括：
图像获得模块，用于获得待定位车牌图像；
图像发送模块，用于发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
特征提取模块，用于所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
组合获得模块，用于获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
区域确定模块，用于根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
可选的，所述区域确定模块，包括：
目标区域确定子模块，用于将所获得的组合中置信度大于预设阈值的定位区域确定为目标定位区域；
车牌区域确定子模块，用于根据所述目标定位区域，确定所述待定位车牌图像的车牌区域。
可选的，所述车牌区域确定子模块，具体用于：
确定所述目标定位区域中车牌框区域和车牌字符区域之间的相对位置；
将所述相对位置处于预设的相对位置范围内的目标定位区域，确定为所述待定位车牌图像的车牌区域。
可选的，所述车牌区域确定子模块，具体用于：将所述目标定位区域中的车牌字符区域，确定为所述待定位车牌图像的车牌区域。
可选的，所述装置还包括：
车牌识别模块，用于对确定的车牌区域进行字符识别，获得所述待定位 车牌图像的车牌号码。
为了达到上述目的，另一方面，本申请实施例公开一种电子设备，适用于车牌定位，所述电子设备包括：
壳体、处理器、存储器、电路板和电源电路，其中，电路板安置在壳体围成的空间内部，处理器和存储器设置在电路板上；电源电路，用于为电子设备的各个电路或器件供电；存储器用于存储可执行程序代码；处理器通过读取存储器中存储的可执行程序代码来运行与可执行程序代码对应的程序，以用于执行以下步骤：
获得待定位车牌图像；
发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
可选的，所述处理器，将所获得的组合中置信度大于预设阈值的定位区域确定为目标定位区域；根据所述目标定位区域，确定所述待定位车牌图像的车牌区域。
可选的，所述处理器，确定所述目标定位区域中车牌框区域和车牌字符区域之间的相对位置；将所述相对位置处于预设的相对位置范围内的目标定位区域，确定为所述待定位车牌图像的车牌区域。
为了达到上述目的，另一方面，本申请实施例公开了一种应用程序，所述应用程序用于在运行时执行本申请实施例提供的车牌定位方法。其中，该 车牌定位方法包括：
获得待定位车牌图像；
发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
为了达到上述目的，另一方面，本申请实施例公开了一种存储介质，用于存储可执行代码，所述可执行代码在运行时用于执行本申请实施例提供的车牌定位方法。其中，该车牌定位方法包括：
获得待定位车牌图像；
发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
由上述技术方案可见，本申请实施例提供的方案中，在获得待定位车牌图像之后，将待定位车牌图像发送至用于车牌定位的目标网络，其中，目标网络包括特征提取层和回归层；然后，特征提取层提取待定位车牌图像的特征值，并发送特征值至回归层，获得回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；最后，根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。其中，目标网络预先通过样本车牌图像训练而成，样本车牌图像包括正样本车牌图像，正样本车牌图像包括车牌框区域和车牌字符区域的组合。
也就是说，本申请实施例可以根据目标网络检测出的车牌框区域和车牌字符区域的组合对应的定位区域，确定待定位车牌图像的车牌区域。由于待定位车牌图像中可能存在很多非车牌字符区域，而根据车牌框区域和车牌字符区域之间的组合关系可以从众多的干扰因素中准确定位车牌区域，无需根据字符与背景的交界点等特征进行定位，因此应用本申请实施例提供的方案，能够提高车牌定位过程的准确性。
附图说明
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单的介绍。显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1a为包含车牌区域的车牌图像示例；
图1b为字符数量较少的车牌示例；
图1c为车身包含非车牌字符的车牌图像示例；
图2为本申请实施例提供的车牌定位方法的一种流程示意图；
图3为几个车牌中车牌框区域和车牌字符区域的组合示例；
图4为目标网络的一种结构示意图；
图5为本申请实施例提供的车牌定位装置的一种结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整的描述。显然，所描述的实施例仅仅是本申请的一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本申请保护的范围。
本申请实施例提供了一种车牌定位方法及装置，应用于电子设备，该电子设备可以是终端设备或服务器等，其中，终端设备可以包括计算机、平板电脑、智能手机、行车记录仪等设备。应用本申请实施例中的技术方案进行车牌定位，能够提高车牌定位过程的准确性。下面通过具体实施例，对本申请进行详细说明。
图2为本申请实施例提供的车牌定位方法的一种流程示意图，应用于电子设备。该方法具体包括如下步骤：
步骤S201：获得待定位车牌图像。
具体的，步骤S201是由执行主体执行的步骤。需要说明的是，作为执行主体的电子设备内部可以包含图像采集设备，也可以不包含图像采集设备。
具体的，当作为执行主体的电子设备内部包含图像采集设备时，电子设备在获得待定位车牌图像时，可以包括：接收图像采集设备采集的待定位车牌图像。
当作为执行主体的电子设备内部不包含图像采集设备时，该电子设备可以与外部的图像采集设备相连，电子设备在获得待定位车牌图像时，可以包括：获取图像采集设备采集的待定位车牌图像。并且，获取的待定位车牌图像可以是图像采集设备实时采集的，也可以不是实时采集的，而是图像采集设备预先采集好之后存储起来的。
上述待定位车牌图像可以理解为，要进行车牌区域定位的图像。可以理解的是，车牌通常是安装或放置于车辆上的，因此，上述待定位车牌图像可以理解为：要进行车牌区域定位的包含车辆的图像。基于此，上述待定位车牌图像可以是道路上抓拍的包含车辆的图像，可以是在停车场拍摄的包含车 辆的图像，等等。当然，上述待定位车牌图像可以是任意一种包含车牌区域的图像，本申请对待定位车牌图像的获得方式不做具体限定。
步骤S202：发送所述待定位车牌图像至用于车牌定位的目标网络。
具体的，步骤S202是由执行主体执行的步骤。其中，所述目标网络包括特征提取层和回归层。所述特征提取层，用于提取所述待定位车牌图像的特征值，并将所述特征值输入所述回归层。所述回归层，用于根据所述特征值得到车牌框区域和车牌字符区域的组合。所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合。
上述车牌框区域可以理解为，车牌外边框所围成的区域。车牌字符区域可以理解为，车牌区域中车牌字符所在的区域。例如，图3中有4个车牌，每个车牌中编号为1的区域为车牌框区域，编号为2的区域为车牌字符区域。
具体的，发送所述待定位车牌图像至用于车牌定位的目标网络，可以包括：发送所述待定位车牌图像至用于车牌定位的目标网络中的特征提取层。
本实施例中，目标网络可以是卷积神经网络等深度学习网络。目标网络还可以包括输入层，该输入层可以用于接收待定位车牌图像，对待定位车牌图像进行预处理，并将预处理后的待定位车牌图像输入特征提取层。所述预处理具体可以包括以下情况中的至少一种：对图像的像素值进行归一化；按照预设灰度阈值，对图像像素的灰度值进行处理，以便改善图像过亮或过暗的情况；对图像的尺寸进行修改等。
当目标网络包括输入层时，发送所述待定位车牌图像至用于车牌定位的目标网络，可以包括：发送所述待定位车牌图像至用于车牌定位的目标网络中的输入层。
作为一种具体实施方式，目标网络还可以包括输出层。当回归层得到车牌框区域和车牌字符区域的组合时，可以将所述组合通过输出层进行输出。另外，输出层也可以对上述组合进行一定处理后再输出，例如数据封装等。
作为一种具体实施方式，上述目标网络可以同时包含输入层、特征提取层、回归层、输出层等。当上述目标网络为卷积神经网络时，该目标网络也 可以称为级联卷积神经网络。应用于级联卷积神经网络中的变化卷积核和激活函数，可以提取更复杂的字符特征，加大字符特征和非字符特征之间的类间差，同时也减小字符特征和非字符特征之间的类内差，从而更容易使回归层回归出车牌字符区域。
作为一个例子，图4给出了目标网络的一种结构示意图，其中，目标网络包括：输入层1、特征提取层2、回归层3和输出层4。在进行车牌定位时，将待定位车牌图像输入输入层1，从输出层4获得针对待定位车牌图像的车牌框区域和车牌字符区域的组合，该组合即为输出层的输出结果。
需要说明的是，输出结果中可以包含车牌框区域和车牌字符区域的组合对应的定位区域及对应的置信度。当然，输出结果中可能包含一个或多个定位区域。
本实施例中，目标网络是预先训练得到的。为了提高所训练的目标网络的鲁棒性，在训练模型时，样本车牌图像中还可以包括负样本车牌图像，该负样本车牌图像可以理解为不包含车牌区域的图像，即不包含车牌框区域和车牌字符区域的图像。
具体的，训练目标网络时，可以预先获得大量包含正样本车牌图像和负样本车牌图像的样本车牌图像，其中，正样本车牌图像中携带车牌框区域和车牌字符区域的组合标记；并发送样本车牌图像至特征提取层。特征提取层接收到样本车牌图像后，获得样本车牌图像的特征值，并发送特征值以及样本车牌图像至回归层。回归层接收特征提取层发送的信息，并根据该信息，从样本车牌图像中确定疑似定位区域，然后计算疑似定位区域与标准定位区域的重合度，所述标准定位区域为所标记的车牌框区域和车牌字符区域的组合对应的定位区域。然后，根据比对结果不断调整回归层的相关参数。当疑似定位区域与标准定位区域的重合度高于设置阈值时，便可以认为目标网络训练完成。
作为一种具体实施方式，当车牌区域包括主字段区域和从字段区域时，可以将主字段区域选择为车牌字符区域进行标记。上述主字段可以理解为具有以下特点的字符部分：字符尺寸较大，分布在车牌中间区域。上述从字段可以理解为具有以下特点的字符部分：字符尺寸比主字段字符小，分布在车 牌边缘处，具有标识意义(标识国家或地区特点)。
例如，在图3所示车牌中，第一个车牌中包括两个字符“5”，左侧的字符“5”表示车辆归属的地区，为从字段；右侧的字符“5”表示车牌的号码，为主字段。第三个车牌中字符“36”为主字段，字符“N”为从字段。
步骤S203：所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层。
具体的，步骤S203是由特征提取层执行的步骤。
具体的，获得特征提取层提取的所述待定位车牌图像的特征值，可以包括：获得特征提取层提取的针对所述待定位车牌图像中像素的特征值。
作为一种具体实施方式，当目标网络包含输入层时，步骤S203可以包括以下实施方式：
方式一，获得所述输入层预处理后的待定位车牌图像，并发送所述预处理后的待定位车牌图像至特征提取层，获取特征提取层提取的预处理后的待定位车牌图像的特征值，并发送所述特征值至回归层。在该实施方式中，输入层对待定位车牌图像进行预处理，并发送预处理后的待定位车牌图像至特征提取层。
方式二，获得所述输入层发送的待定位车牌图像，并发送所述待定位车牌图像至特征提取层，获取特征提取层提取的待定位车牌图像的特征值，并发送所述特征值至回归层。在该实施方式中，输入层不对待定位车牌图像进行预处理，而是直接发送待定位车牌图像至特征提取层。
步骤S204：获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合。
具体的，步骤S204是由执行主体执行的步骤。
可以理解的是，回归层在根据特征值得到车牌框区域和车牌字符区域的组合时，也会得到每个组合对应的置信度。因此，上述所获得的组合中还可以包含每个组合的置信度。
作为一种具体实施方式，当目标网络包含输出层时，在步骤S204之后还 可以包括：
发送所获得的组合至输出层，以使所述输出层输出所获得的组合。
当然，在输出层输出所获得的组合时，也可以包括：将所获得的组合按照预设规则进行处理，并输出处理后的组合。预设规则可以包括按照预设格式进行排序等规则。
步骤S205：根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
具体的，步骤S205是由执行主体执行的步骤，步骤S205可以包括：将所获得的组合中置信度大于预设阈值的定位区域确定为目标定位区域，根据所述目标定位区域，确定所述待定位车牌图像的车牌区域。
作为确定目标定位区域的一个例子，已知所获得的组合包括组合1、组合2、组合3和组合4，每个组合均为车牌框区域和车牌字符区域的组合，并且这四个组合的置信度分别为320、270、950和890。已知预设阈值为800，则确定组合3和组合4的置信度均大于预设阈值，可以将组合3和组合4对应的定位区域确定为目标定位区域。
进一步的，根据所述目标定位区域，确定所述待定位车牌图像的车牌区域，可以包括以下实施方式：
方式一，确定所述目标定位区域中车牌框区域和车牌字符区域之间的相对位置，将所述相对位置处于预设的相对位置范围内的目标定位区域，确定为所述待定位车牌图像的车牌区域。
作为一种实施方式，车牌框区域和车牌字符区域之间的相对位置，可以采用两区域对应边界之间的距离表示，也可以采用两区域中心点的相对位置并结合两区域的长度比值和宽度比值来表示。
需要说明的是，相对位置范围可以根据预先采集的正样本车牌图像中车牌框区域和车牌字符区域之间的相对位置确定。可以理解的是，根据预设的相对位置范围对输出结果进行筛选，可以有效地从所获得的组合中滤除非车牌区域，提高车牌定位过程的准确性。
方式二，将所述目标定位区域中的车牌字符区域，确定为所述待定位车牌图像的车牌区域。
可以理解的是，车牌框区域比车牌字符区域包含更多的图像信息，同时也包含很多除车牌字符之外的干扰信息。因此，为了提高车牌定位过程的准确性，可以直接将目标定位区域中的车牌字符区域，确定为所述待定位车牌图像的车牌区域。
当然，在本实施例中，还可以进一步针对目标定位区域中的车牌框区域，确定车牌的从字段部分。具体的，可以由回归层的分支进行车牌类别判断，即根据车牌框内整体纹理信息确定车牌类别，再由预设的车牌类别确定从字段与主字段的位置关系，从而从车牌框区域中确定车牌的从字段部分。
由上述内容可知，本实施例提供的方案中，在获得待定位车牌图像之后，将待定位车牌图像发送至用于车牌定位的目标网络，其中，目标网络包括特征提取层和回归层；然后，特征提取层提取待定位车牌图像的特征值，并发送特征值至回归层，获得回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；最后，根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。其中，目标网络预先通过样本车牌图像训练而成，样本车牌图像包括正样本车牌图像，正样本车牌图像包括车牌框区域和车牌字符区域的组合。
也就是说，本实施例根据目标网络检测出的车牌框区域和车牌字符区域的组合对应的定位区域，确定待定位车牌图像的车牌区域。由于待定位车牌图像中可能存在很多非车牌字符区域或非车牌边框区域，例如，图1c所示的车牌图像中，车身上车牌区域以外还存在一些其他文字部分和区域框部分等干扰因素；而根据车牌框区域和车牌字符区域之间的组合关系可以从众多的干扰因素中准确定位车牌区域，无需根据字符与背景的交界点等特征进行定位，因此应用本实施例的方案，能够提高车牌定位过程的准确性。
基于图2所示实施例的一种具体实施方式中，在确定所述待定位车牌图像的车牌区域之后，所述方法还可以包括：对确定的车牌区域进行字符识别，获得所述待定位车牌图像的车牌号码。
图5为本申请实施例提供的车牌定位装置的一种结构示意图，与图2所示方法实施例相对应，应用于电子设备，所述装置包括：
图像获得模块501，用于获得待定位车牌图像；
图像发送模块502，用于发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
特征提取模块503，用于所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
组合获得模块504，用于获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
区域确定模块505，用于根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
基于图5所示实施例的一种具体实施方式中，所述区域确定模块505可以包括：
目标区域确定子模块(图中未示出)，用于将所获得的组合中置信度大于预设阈值的定位区域确定为目标定位区域；
车牌区域确定子模块(图中未示出)，用于根据所述目标定位区域，确定所述待定位车牌图像的车牌区域。
基于图5所示实施例的一种具体实施方式中，所述车牌区域确定子模块，具体可以用于：
确定所述目标定位区域中车牌框区域和车牌字符区域之间的相对位置，将所述相对位置处于预设的相对位置范围内的目标定位区域，确定为所述待定位车牌图像的车牌区域。
基于图5所示实施例的一种具体实施方式中，所述车牌区域确定子模块，具体用于：
将所述目标定位区域中的车牌字符区域，确定为所述待定位车牌图像的车牌区域。
基于图5所示实施例的一种具体实施方式中，所述装置还可以包括：车牌识别模块(图中未示出)，用于对确定的车牌区域进行字符识别，获得所述待定位车牌图像的车牌号码。
由于上述装置实施例是基于方法实施例得到的，与该方法具有相同的技术效果，因此装置实施例的技术效果在此不再赘述。
对于装置实施例而言，由于其基本相似于方法实施例，所以描述得比较简单，相关之处参见方法实施例的部分说明即可。
本申请实施例提供了一种电子设备，适用于车牌定位，所述电子设备包括：
壳体、处理器、存储器、电路板和电源电路，其中，电路板安置在壳体围成的空间内部，处理器和存储器设置在电路板上；电源电路，用于为电子设备的各个电路或器件供电；存储器用于存储可执行程序代码；处理器通过读取存储器中存储的可执行程序代码来运行与可执行程序代码对应的程序，以用于执行以下步骤：
获得待定位车牌图像；
发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
其中，该电子设备可以以多种形式存在，包括但不限于：
(1)移动通信设备：这类设备的特点是具备移动通信功能，并且以提供话音、数据通信为主要目标。这类终端包括：智能手机(例如iPhone)、多媒体手机、功能性手机以及低端手机等。
(2)超移动个人计算机设备：这类设备属于个人计算机的范畴，有计算和处理功能，一般也具备移动上网特性。这类终端包括：PDA、MID和UMPC设备等，例如iPad。
(3)便携式娱乐设备：这类设备可以显示和播放多媒体内容。该类设备包括：音频、视频播放器(例如iPod)，掌上游戏机，电子书，以及智能玩具和便携式车载导航设备。
(4)服务器：提供计算服务的设备，服务器的构成包括处理器、硬盘、内存、***总线等，服务器和通用的计算机架构类似，但是由于需要提供高可靠的服务，因此在处理能力、稳定性、可靠性、安全性、可扩展性、可管理性等方面要求较高。
(5)其他具有数据交互功能的电子装置。
可见，本实施例可以根据目标网络检测出的车牌框区域和车牌字符区域的组合对应的定位区域，确定待定位车牌图像的车牌区域。由于待定位车牌图像中可能存在很多非车牌字符区域，而根据车牌框区域和车牌字符区域之间的组合关系可以从众多的干扰因素中准确定位车牌区域，无需根据字符与背景的交界点等特征进行定位，因此应用本实施例提供的方案，能够提高车牌定位过程的准确性。
在本实施例的一种具体实施方式中，所述处理器，将所获得的组合中置信度大于预设阈值的定位区域确定为目标定位区域；根据所述目标定位区域，确定所述待定位车牌图像的车牌区域。
在本实施例的一种具体实施方式中，所述处理器，确定所述目标定位区域中车牌框区域和车牌字符区域之间的相对位置；将所述相对位置处于预设的相对位置范围内的目标定位区域，确定为所述待定位车牌图像的车牌区域。
本申请实施例提供了一种应用程序，所述应用程序用于在运行时执行本申请实施例提供的车牌定位方法。其中，该车牌定位方法包括：
获得待定位车牌图像；
发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
可见，本实施例可以根据目标网络检测出的车牌框区域和车牌字符区域的组合对应的定位区域，确定待定位车牌图像的车牌区域。由于待定位车牌图像中可能存在很多非车牌字符区域，而根据车牌框区域和车牌字符区域之间的组合关系可以从众多的干扰因素中准确定位车牌区域，无需根据字符与背景的交界点等特征进行定位，因此应用本实施例提供的方案，能够提高车牌定位过程的准确性。
本申请实施例提供了一种存储介质，用于存储可执行代码，所述可执行代码在运行时用于执行本申请实施例提供的车牌定位方法。其中，该车牌定位方法包括：
获得待定位车牌图像；
发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
可见，本实施例可以根据目标网络检测出的车牌框区域和车牌字符区域的组合对应的定位区域，确定待定位车牌图像的车牌区域。由于待定位车牌图像中可能存在很多非车牌字符区域，而根据车牌框区域和车牌字符区域之间的组合关系可以从众多的干扰因素中准确定位车牌区域，无需根据字符与背景的交界点等特征进行定位，因此应用本实施例提供的方案，能够提高车牌定位过程的准确性。
需要说明的是，在本文中，诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
本领域普通技术人员可以理解，上述实施方式中的全部或部分步骤是能够通过程序指令相关的硬件来完成的，所述的程序可以存储于计算机可读取存储介质中。这里所称存储介质，是指ROM/RAM、磁碟、光盘等。
以上所述仅为本申请的较佳实施例而已，并非用于限定本申请的保护范围。凡在本申请的精神和原则之内所做的任何修改、等同替换、改进等，均包含在本申请的保护范围内。

Claims (15)

  1. 一种车牌定位方法，其特征在于，所述方法包括：
    获得待定位车牌图像；
    发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
    所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
    获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
    根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域，包括：
    将所获得的组合中置信度大于预设阈值的定位区域确定为目标定位区域；
    根据所述目标定位区域，确定所述待定位车牌图像的车牌区域。
  3. 根据权利要求2所述的方法，其特征在于，所述根据所述目标定位区域，确定所述待定位车牌图像的车牌区域，包括：
    确定所述目标定位区域中车牌框区域和车牌字符区域之间的相对位置；
    将所述相对位置处于预设的相对位置范围内的目标定位区域，确定为所述待定位车牌图像的车牌区域。
  4. 根据权利要求2所述的方法，其特征在于，所述根据所述目标定位区域，确定所述待定位车牌图像的车牌区域，包括：
    将所述目标定位区域中的车牌字符区域，确定为所述待定位车牌图像的 车牌区域。
  5. 根据权利要求1-4任一项所述的方法，其特征在于，所述方法还包括：
    对确定的车牌区域进行字符识别，获得所述待定位车牌图像的车牌号码。
  6. 一种车牌定位装置，其特征在于，所述装置包括：
    图像获得模块，用于获得待定位车牌图像；
    图像发送模块，用于发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
    特征提取模块，用于所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
    组合获得模块，用于获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
    区域确定模块，用于根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
  7. 根据权利要求6所述的装置，其特征在于，所述区域确定模块，包括：
    目标区域确定子模块，用于将所获得的组合中置信度大于预设阈值的定位区域确定为目标定位区域；
    车牌区域确定子模块，用于根据所述目标定位区域，确定所述待定位车牌图像的车牌区域。
  8. 根据权利要求7所述的装置，其特征在于，所述车牌区域确定子模块，具体用于：
    确定所述目标定位区域中车牌框区域和车牌字符区域之间的相对位置；
    将所述相对位置处于预设的相对位置范围内的目标定位区域，确定为所述待定位车牌图像的车牌区域。
  9. 根据权利要求7所述的装置，其特征在于，所述车牌区域确定子模块，具体用于：
    将所述目标定位区域中的车牌字符区域，确定为所述待定位车牌图像的车牌区域。
  10. 根据权利要求6-9任一项所述的装置，其特征在于，所述装置还包括：
    车牌识别模块，用于对确定的车牌区域进行字符识别，获得所述待定位车牌图像的车牌号码。
  11. 一种电子设备，其特征在于，适用于车牌定位，所述电子设备包括：
    壳体、处理器、存储器、电路板和电源电路，其中，电路板安置在壳体围成的空间内部，处理器和存储器设置在电路板上；电源电路，用于为电子设备的各个电路或器件供电；存储器用于存储可执行程序代码；处理器通过读取存储器中存储的可执行程序代码来运行与可执行程序代码对应的程序，以用于执行以下步骤：
    获得待定位车牌图像；
    发送所述待定位车牌图像至用于车牌定位的目标网络，其中，所述目标网络包括特征提取层和回归层，所述目标网络预先通过样本车牌图像训练而成，所述样本车牌图像包括正样本车牌图像，所述正样本车牌图像包括车牌框区域和车牌字符区域的组合；
    所述特征提取层提取所述待定位车牌图像的特征值，并发送所述特征值至所述回归层；
    获得所述回归层根据所述特征值得到的车牌框区域和车牌字符区域的组合；
    根据所获得的组合对应的定位区域，确定所述待定位车牌图像的车牌区域。
  12. 根据权利要求11所述的电子设备，其特征在于，所述处理器，将所获得的组合中置信度大于预设阈值的定位区域确定为目标定位区域；根据所述目标定位区域，确定所述待定位车牌图像的车牌区域。
  13. 根据权利要求12所述的电子设备，其特征在于，所述处理器，确定所述目标定位区域中车牌框区域和车牌字符区域之间的相对位置；将所述相对位置处于预设的相对位置范围内的目标定位区域，确定为所述待定位车牌图像的车牌区域。
  14. 一种应用程序，其特征在于，所述应用程序用于在运行时执行权利要求1-5任一项所述的车牌定位方法。
  15. 一种存储介质，其特征在于，用于存储可执行代码，所述可执行代码在运行时用于执行权利要求1-5任一项所述的车牌定位方法。
PCT/CN2017/106661 2016-12-30 2017-10-18 一种车牌定位方法及装置 WO2018121006A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP17889416.8A EP3564855A4 (en) 2016-12-30 2017-10-18 LICENSE PLATE POSITIONING METHOD AND DEVICE
US16/473,787 US11126882B2 (en) 2016-12-30 2017-10-18 Method and device for license plate positioning

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201611261487.4A CN108268867B (zh) 2016-12-30 2016-12-30 一种车牌定位方法及装置
CN201611261487.4 2016-12-30

Publications (1)

Publication Number Publication Date
WO2018121006A1 true WO2018121006A1 (zh) 2018-07-05

Family

ID=62706824

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/106661 WO2018121006A1 (zh) 2016-12-30 2017-10-18 一种车牌定位方法及装置

Country Status (4)

Country Link
US (1) US11126882B2 (zh)
EP (1) EP3564855A4 (zh)
CN (1) CN108268867B (zh)
WO (1) WO2018121006A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110519485A (zh) * 2019-09-09 2019-11-29 Oppo广东移动通信有限公司 图像处理方法、装置、存储介质及电子设备
CN111027555A (zh) * 2018-10-09 2020-04-17 杭州海康威视数字技术股份有限公司 一种车牌识别方法、装置及电子设备
CN112348044A (zh) * 2019-08-09 2021-02-09 上海高德威智能交通***有限公司 车牌检测方法、装置及设备

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3599572B1 (en) * 2018-07-27 2021-09-01 JENOPTIK Traffic Solutions UK Ltd Method and apparatus for recognizing a license plate of a vehicle
CN109190687A (zh) * 2018-08-16 2019-01-11 新智数字科技有限公司 一种神经网络***及其识别车辆属性的方法
US10963719B1 (en) * 2019-03-04 2021-03-30 Ccc Information Services Inc. Optimized vehicle license plate recognition
KR20200119409A (ko) * 2019-03-28 2020-10-20 한국전자통신연구원 번호판 판독을 위한 장치 및 방법
CN112115748B (zh) * 2019-06-21 2023-08-25 腾讯科技（深圳）有限公司 证件图像识别方法、装置、终端及存储介质
CN111079744B (zh) * 2019-12-06 2020-09-01 鲁东大学 适用于复杂光照环境的车辆车牌智能识别方法及装置
CN112348020B (zh) * 2020-12-03 2023-10-20 北京智芯原动科技有限公司 基于特征图的贝塞尔车牌对齐方法及装置
CN113177552B (zh) * 2021-05-27 2024-04-26 的卢技术有限公司 一种基于深度学习的车牌识别方法
CN114782257B (zh) * 2022-06-22 2022-11-15 深圳市爱深盈通信息技术有限公司 车牌拼接方法、装置、识别方法、设备终端和存储介质

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103699905A (zh) * 2013-12-27 2014-04-02 深圳市捷顺科技实业股份有限公司 一种车牌定位方法及装置
CN103870803A (zh) * 2013-10-21 2014-06-18 北京邮电大学 一种基于粗定位与精定位融合的车牌识别方法和***

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9158995B2 (en) * 2013-03-14 2015-10-13 Xerox Corporation Data driven localization using task-dependent representations
CN103530600B (zh) * 2013-06-06 2016-08-24 东软集团股份有限公司 复杂光照下的车牌识别方法及***
CN103902981A (zh) * 2014-04-02 2014-07-02 浙江师范大学 一种基于字符融合特征的车牌字符识别方法及***
CN105373794B (zh) * 2015-12-14 2018-02-06 河北工业大学 一种车牌识别方法

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103870803A (zh) * 2013-10-21 2014-06-18 北京邮电大学 一种基于粗定位与精定位融合的车牌识别方法和***
CN103699905A (zh) * 2013-12-27 2014-04-02 深圳市捷顺科技实业股份有限公司 一种车牌定位方法及装置

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
GAO,WEI: "Research Vehicle License Plate Localization in Complex Scene and Its Application", CHINESE MASTER'S THESES, no. 10, 15 October 2016 (2016-10-15), pages 1 - 73, XP009515485, ISSN: 1674-0246 *
LUO, BIN ,ET AL.: "Learning corner Regression-based Fully Convolutional Neutral network for License Plate Localization in Complex Scene", JOURNAL OF DATA ADQUISITION & PROCESSING, no. 1, 15 January 2016 (2016-01-15), pages 65 - 71, XP009515372, ISSN: 1004-9037 *
See also references of EP3564855A4

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111027555A (zh) * 2018-10-09 2020-04-17 杭州海康威视数字技术股份有限公司 一种车牌识别方法、装置及电子设备
CN111027555B (zh) * 2018-10-09 2023-09-26 杭州海康威视数字技术股份有限公司 一种车牌识别方法、装置及电子设备
CN112348044A (zh) * 2019-08-09 2021-02-09 上海高德威智能交通***有限公司 车牌检测方法、装置及设备
CN110519485A (zh) * 2019-09-09 2019-11-29 Oppo广东移动通信有限公司 图像处理方法、装置、存储介质及电子设备
CN110519485B (zh) * 2019-09-09 2021-08-31 Oppo广东移动通信有限公司 图像处理方法、装置、存储介质及电子设备

Also Published As

Publication number Publication date
CN108268867A (zh) 2018-07-10
US11126882B2 (en) 2021-09-21
CN108268867B (zh) 2020-10-30
EP3564855A1 (en) 2019-11-06
US20200125876A1 (en) 2020-04-23
EP3564855A4 (en) 2020-01-08

Similar Documents

Publication Publication Date Title
WO2018121006A1 (zh) 一种车牌定位方法及装置
WO2018090771A1 (zh) 一种车牌识别方法及装置
US10699103B2 (en) Living body detecting method and apparatus, device and storage medium
US11830230B2 (en) Living body detection method based on facial recognition, and electronic device and storage medium
CN109086669B (zh) 人脸识别身份验证方法、装置、电子设备
CN109886308B (zh) 一种基于目标级别的双传感器数据融合方法和装置
US9436883B2 (en) Collaborative text detection and recognition
WO2019071664A1 (zh) 结合深度信息的人脸识别方法、装置及存储介质
WO2018028306A1 (zh) 一种车牌号码识别方法及装置
US10769476B2 (en) License plate detection method and device
KR20140010164A (ko) 객체에서 텍스트 정보를 인식하기 위한 시스템 및 방법
CN111368811B (zh) 活体检测方法、装置、设备及存储介质
CN111507324B (zh) 卡片边框识别方法、装置、设备和计算机存储介质
US9679218B2 (en) Method and apparatus for image matching
US20220327861A1 (en) Method for recognizing masked faces, device for recognizing masked faces, and computer storage medium
CN114299546A (zh) 识别宠物身份的方法、装置、存储介质及电子设备
CN112380978B (zh) 基于关键点定位的多人脸检测方法、***及存储介质
CN116152870A (zh) 人脸识别方法、装置、电子设备及计算机可读存储介质
CN112257628A (zh) 一种户外比赛运动员的身份识别方法、装置及设备
CN112990009A (zh) 基于端到端的车道线检测方法、装置、设备及存储介质
CN110751163B (zh) 目标定位方法及其装置、计算机可读存储介质和电子设备
CN114067394A (zh) 人脸活体检测方法、装置、电子设备及存储介质
CN108694347B (zh) 图像处理方法和装置
Ahmed GPark: Vehicle parking management system using smart glass
CN112115740A (zh) 用于处理图像的方法和装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17889416

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2017889416

Country of ref document: EP

Effective date: 20190730