Vehicle abnormal gathering monitoring method, device, computer equipment and storage medium
By automatically identifying vehicle type and car owner information, combining abnormal aggregation blacklist, calculating the number and proportion of vehicles, the misjudgment problem of existing vehicle abnormal aggregation monitoring methods is solved, the accuracy and efficiency of monitoring are improved, and the vehicle abnormality is achieved. Automatic identification of aggregation situations.
Patent Information
- Application Number
- CN202111660556.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-30
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2041-12-30
AI Technical Summary
The existing vehicle abnormal aggregation monitoring methods are prone to misjudgment, resulting in low accuracy and accuracy of early warnings, wasted resources, and manual identification of whether image data is aggregation, which is costly and energy-consuming.
By obtaining image data within the preset time period, we automatically identify specific vehicle types and car owner information, calculate the number of vehicles and car owner proportion based on the abnormal aggregation blacklist, determine whether there is abnormal aggregation, and realize automated monitoring.
It improves the accuracy and accuracy of vehicle abnormal aggregation monitoring, saves labor costs and energy, and realizes objective, accurate and automatic identification of vehicle abnormal aggregation conditions.
Smart Images

Figure CN114494994B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device, computer equipment and storage medium for monitoring abnormal vehicle aggregation. Background Art
[0002] At present, with the progress of society and the rapid development of economy, people's quality of life has been improved, and the use and mobility of vehicles have also increased rapidly. It is easy to have abnormal gathering of people or vehicles, such as collective fighting and other illegal acts. The alarm call received by the relevant public security departments is often delayed after the situation occurs, which is easy to cause serious dangerous incidents. In the existing warning / monitoring methods of abnormal gathering, most of them use the positioning information of the vehicle-mounted GPS terminal to determine whether there is a high-density gathering of vehicles in a certain area, and then conduct warning / monitoring according to the density. However, this solution is prone to misjudgment, such as rush hour or an exhibition or the gathered vehicles are not the objects of concern (also known as the owners of illegal behaviors or vehicles with illegal records), so that the accuracy and correctness of the warning / monitoring of abnormal gatherings are not high, resulting in a waste of resources for staff to handle matters related to abnormal gatherings warned / monitored. Summary of the invention
[0003] The present invention provides a method, device, computer equipment and storage medium for monitoring abnormal vehicle aggregation, which can determine the number of target vehicles of a specific vehicle type and the information of their owners from massive image data, and objectively, accurately and automatically identify whether there is abnormal aggregation in the abnormal aggregation detection results of the monitored area based on an abnormal aggregation blacklist.
[0004] A method for monitoring abnormal gathering of vehicles, comprising:
[0005] Acquire multiple image data captured in the area to be monitored within a preset time period;
[0006] Determine the number of target vehicles of a specific vehicle type that appear in the monitored area within a preset time period and the owner information of the target vehicles based on the plurality of image data;
[0007] Based on the number of vehicles, the owner information of the target vehicle and a preset abnormal aggregation blacklist, determining a percentage of owners with abnormal aggregation records;
[0008] According to the number of vehicles and the proportion of vehicle owners, the abnormal aggregation detection result of the monitored area within a preset time period is determined.
[0009] A vehicle abnormal gathering monitoring device, comprising:
[0010] An acquisition module is used to acquire a plurality of image data captured in the monitored area within a preset time period;
[0011] A statistical module, used to determine the number of target vehicles of a specific vehicle type appearing in the monitored area within a preset time period and the owner information of the target vehicles according to the plurality of image data;
[0012] A determination module, configured to determine a percentage of vehicle owners having abnormal aggregation records based on the number of vehicles, the owner information of the target vehicle, and a preset abnormal aggregation blacklist;
[0013] The output module is used to determine the abnormal aggregation detection result of the monitored area within a preset time period according to the number of vehicles and the proportion of vehicle owners.
[0014] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned vehicle abnormal aggregation monitoring method when executing the computer program.
[0015] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned vehicle abnormal aggregation monitoring method.
[0016] The method, device, computer equipment and storage medium for monitoring abnormal gathering of vehicles provided by the present invention obtain a plurality of image data captured in the area to be monitored within a preset time period; determine the number of target vehicles of a specific vehicle type appearing in the area to be monitored within the preset time period and the owner information of the target vehicles based on the plurality of image data; determine the proportion of vehicle owners with abnormal gathering records based on the number of vehicles, the owner information of the target vehicles and a preset abnormal gathering blacklist; determine the abnormal gathering detection result of the area to be monitored within the preset time period based on the number of vehicles and the proportion of vehicle owners. Therefore, the present invention obtains image data captured in the area to be monitored within the preset time period, It can also automatically determine the number of target vehicles of a specific vehicle type and the information of vehicle owners from all image data, and automatically determine the proportion of vehicle owners based on the abnormal aggregation blacklist. It can also automatically output the abnormal aggregation detection result of the monitored area based on the comprehensive value of the number of vehicles and the proportion of vehicle owners. In this way, it can obtain image data from a large amount of captured data, and determine the number of target vehicles of a specific vehicle type and the information of vehicle owners. Based on the abnormal aggregation blacklist, it can objectively, accurately and automatically identify whether there is abnormal aggregation in the abnormal aggregation detection result of the monitored area. There is no need to manually identify whether there is aggregation in the image data, which greatly saves labor costs and energy, and improves the accuracy and correctness of abnormal vehicle aggregation monitoring. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.
[0018] Figure 1 Schematic diagram of the application environment of the abnormal vehicle aggregation monitoring method in one embodiment of the present invention;
[0019] Figure 2 is a flow chart of a method for monitoring abnormal aggregation of vehicles in one embodiment of the present invention;
[0020] Figure 3 is a flow chart of step S10 of a method for monitoring abnormal aggregation of vehicles in one embodiment of the present invention;
[0021] Figure 4 is a principle block diagram of a vehicle abnormal gathering monitoring device in one embodiment of the present invention;
[0022] Figure 5 is a schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION
[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0024] The vehicle abnormal gathering monitoring method provided by the present invention can be applied in the following aspects: Figure 1 In an application environment, a client (computer device or terminal) communicates with a server via a network. The client (computer device or terminal) includes but is not limited to various personal computers, laptops, smart phones, tablet computers, and portable wearable devices. The server can be implemented as an independent server or a server cluster consisting of multiple servers.
[0025] In one embodiment, if Figure 2 As shown, a method for monitoring abnormal aggregation of vehicles is provided, and its technical solution mainly includes the following steps S10-S40:
[0026] S10, obtaining a plurality of image data captured in the area to be monitored within a preset time period.
[0027] It can be understood that the image data is data containing vehicle-related information extracted from the raw data collected in the area to be monitored within a preset time period. The preset time period can be set according to demand, such as 14:00 to 14:59 in 24 hours a day. The area to be monitored is the area to be monitored. The monitored area can be photographed by the deployed image acquisition device. The raw data is the image taken by the image acquisition device of the area to be monitored that has been deployed on the big data platform and the data related to the shooting. The image taken in the raw data may or may not contain a vehicle. The raw data includes a device identifier and a captured image. The device identifier is a unique identification code of the image acquisition device currently taking the image. One device identifier is associated with a monitoring area identifier. The area to be monitored photographed by the image acquisition device can be obtained through the device identifier. Each area to be monitored is assigned a unique identifier. The image acquisition device can be a camera or a timed acquisition device that takes images at preset intervals. The captured image is a frame of image captured by the image acquisition device. The process of extracting image data that needs to be monitored (including vehicle-related information) from massive raw data can be as follows: first, multiple raw data of the area to be monitored within a preset time period are obtained; second, vehicle recognition is performed on the captured images in each raw data, and the first recognition result of which raw data contains the vehicle is identified, and the raw data containing the vehicle is filtered out according to the identified first recognition result; finally, the captured image of the vehicle is extracted from the captured images in the filtered raw data, and the vehicle sub-attribute is identified on the extracted captured image of the vehicle, and the second recognition result of the vehicle focus attribute is identified, and the first recognition result and the second recognition result of the raw data are recorded as the image data of the raw data. The vehicle focus attribute can be set according to the needs. For example, the vehicle focus attribute includes the vehicle license plate, vehicle model (for example: taxi, family car, transportation vehicle, etc.), etc.
[0028] In one embodiment, if Figure 3 As shown, in step S10, a plurality of image data captured in the monitored area within a preset time period is obtained, including:
[0029] S101, obtaining a plurality of original data.
[0030] It can be understood that the original data is the images taken by the image acquisition device of the area to be monitored that has been deployed on the big data platform and the data related to the shooting. The images taken in the original data may contain vehicles or may not contain vehicles. The original data also includes the monitoring time. The monitoring time is the time when the corresponding image acquisition device takes the photos. The original data within the preset time period can be obtained through the monitoring time. The device identifier is the unique identification code of the image acquisition device currently taking the photos. One device identifier is associated with a monitoring area identifier. The area to be monitored taken by the image acquisition device can be obtained through the device identifier. Each area to be monitored is given a unique identifier. The image acquisition device can be a camera or a preset time interval. A timing collector is used to capture images at a certain time. The captured image is a frame of image captured by the image acquisition device. A monitoring area identifier of a monitored area is associated with a risk level. The risk level reflects the degree of risk of aggregation in the monitored area. The risk level will be continuously updated with the passage of time. For example, the risk level is determined by accumulating the number of abnormal aggregation detection results of abnormal aggregation. That is, the risk level of the monitored area will be analyzed with the relevant historical data of the monitored area. The original data is obtained after the collected video is framed by the video acquisition device. The frame processing is the process of extracting each frame of the collected video and marking the monitoring time and device identifier for each frame.
[0031] S102, for each original data, performing vehicle recognition on the captured image in the original data to obtain a first recognition result corresponding to the original data; the first recognition result reflects whether the captured image contains a vehicle.
[0032] Understandably, the process of vehicle recognition for the captured images in the original data can be achieved through the trained vehicle recognition network. The vehicle recognition process is to extract vehicle features from the captured images, determine the area containing vehicle features based on the extracted vehicle features, determine the area as a vehicle area, and determine whether the first recognition result of the vehicle is included based on the number of vehicle areas. The vehicle attribute result includes the first recognition result and the recognition results output by other vehicle-related attribute networks. The first recognition result reflects whether the captured image contains a vehicle. The vehicle features are the features of the vehicle, such as: the front end of the vehicle has a license plate feature, the bottom of the vehicle has a tire feature, the front end of the vehicle has a portrait feature, etc. The network structure of the vehicle recognition network can be set according to the needs. For example, the network structure of the vehicle recognition network can be CenterNet, Faster R-CNN, SSD, YOLO and other network structures. Preferably, the network structure of the vehicle recognition network is the network structure of YOLO V3, because the use of YOLO The network structure of V3 can quickly locate the position of the vehicle from different scales. The training process of the vehicle recognition network is to input a sample image containing a vehicle annotation area, extract the vehicle features (convolution) of the sample image, and obtain a small-scale vehicle feature map based on the extracted vehicle features. The small-scale vehicle feature map is upsampled and tensor spliced to obtain a medium-scale vehicle feature map, and then the medium-scale vehicle feature map is upsampled and tensor spliced to obtain a large-scale vehicle feature map.
[0033] Furthermore, multi-scale detection is performed on small-scale, medium-scale and large-scale vehicle feature maps, and a rectangular area with vehicle features is predicted. The loss value between the rectangular area and the vehicle annotation area is calculated, and the parameters of the vehicle recognition network are iteratively updated according to the loss value until the loss value reaches the convergence condition, thereby stopping the training process. The small scale, medium scale and large scale can be 1 / 32 scale, 1 / 16 scale and 1 / 8 scale relative to the size of the captured image, respectively. Vehicle feature maps of different scales can detect more fine-grained vehicle features. Upsampling is the reverse propagation of convolution, that is, the transpose of deconvolution or forward propagation. Tensor splicing is the process of splicing the upsampling results of the intermediate layer and the adjacent next layer in the convolution process. The vehicle process of the vehicle recognition network is to extract vehicle features from the input captured image through the vehicle recognition network, predict the vehicle area based on the extracted vehicle features, and predict the results of a rectangular vehicle area with zero or at least one vehicle. If zero vehicle area is predicted, the first recognition result is determined to contain no vehicle. If at least one vehicle area is predicted, the first recognition result is determined to contain a vehicle, and includes each predicted vehicle area.
[0034] S103, based on the first recognition result corresponding to each original data, all the original data are screened for vehicles, and the original data containing the vehicle is screened out.
[0035] Understandably, according to the first recognition result containing the vehicle, all the original data are screened for the data containing the vehicle, and the original data containing the vehicle are screened out.
[0036] S104, based on the vehicle area in the first recognition result, extract the captured images in the filtered original data to obtain at least one vehicle captured image, and perform vehicle sub-attribute recognition on the vehicle captured image to obtain a second recognition result corresponding to the original data.
[0037] It can be understood that for a filtered original data, according to at least one vehicle area in the first recognition result, the vehicle area is a rectangular range that frames the vehicle in the captured image (the coordinate range in the captured image), and a vehicle captured image corresponding to the vehicle area is copied from the captured image corresponding to the first recognition result, that is, one vehicle area corresponds to one vehicle captured image, and the vehicle sub-attribute recognition is a recognition process of identifying the result of the vehicle focus attribute in the input vehicle captured image through a network of vehicle focus attributes. Other vehicle focus attribute networks include target detection networks, character recognition networks, and vehicle model detection networks. The process of vehicle sub-attribute recognition can be: locating the owner area of the vehicle captured image to obtain the owner area image of the vehicle captured image; and recognizing the license plate and vehicle model attributes of the vehicle captured image to obtain the license plate number and attribute results of the vehicle captured image; finally, the owner area image, license plate number and attribute results of the vehicle captured image are summarized to obtain the results of each sub-attribute of the vehicle, thereby determining the results of each sub-attribute of the vehicle as the second recognition result.
[0038] In one embodiment, the vehicle owner area is located on the captured vehicle image to obtain the vehicle owner area image of the captured vehicle image.
[0039] Understandably, the process of owner area positioning is as follows: first, the input vehicle capture image is preprocessed. The image preprocessing includes image scaling technology and image enhancement technology to make the size of the vehicle capture image consistent with the subsequent positioning of the owner area, and to make the vehicle capture image clearer and the image information contained in it more obvious. After image preprocessing, a to-be-processed image of a preset size (for example, 512×512) is obtained; secondly, the to-be-processed image is subjected to owner area recognition. The process of owner area recognition can be achieved through a trained target detection network. The network structure of the target detection network can be set according to requirements. For example, the network structure of the target detection network can be VGG, CenterNet, or Faster. R-CNN, SSD and YOLO, etc. Preferably, the network structure of the target detection network is a network structure based on VGG19, and the owner's features are extracted from the image to be processed through the target detection network. The owner's features are features related to the owner of the driving seat, such as the owner's features having a face, a steering wheel, and features of a driving state in which the face and the steering wheel are adjacent. Classification and recognition are performed according to the extracted owner's features to identify the detection results of the owner's area range; finally, according to the owner's area range in the detection results, the owner's area image is extracted from the image to be processed.
[0040] Among them, the image scaling technology is to use downsampling and / or bilinear interpolation methods to make the size of an image reach a preset size. The downsampling method is to downsample the vehicle captured image through a preset downsampling unit convolution kernel, and continuously downsample until the image size reaches a preset size. The downsampling method also includes calculating a ratio value based on the preset size and the current size of the vehicle captured image, and then downsampling the current vehicle captured image based on the size of the downsampling convolution kernel corresponding to the ratio value, so that the size of the vehicle captured image that blocks the gun can be reached. The bilinear interpolation method is to divide a target pixel point that needs to be inserted in two directions (X-axis direction and Y-axis direction) A linear interpolation is performed separately, that is, linear interpolation is first used in one direction to obtain a first interpolation, and then linear interpolation is used in another direction to obtain a second interpolation. According to the weighted sum between the first interpolation and the second interpolation, the pixel value of the target pixel point is obtained. After the pixel values of each target pixel are inserted by the bilinear interpolation method, the pixel position of each coordinate point is linearly enlarged according to the linear amplification function between the current size and the preset size, so as to expand it to the preset size. Through the downsampling method and the bilinear interpolation method, the vehicle capture image can be freely scaled. The linear interpolation is to obtain the value by linear calculation of the numerical value based on the two adjacent data on the left and right of the point to be interpolated in the one-dimensional data sequence.
[0041] Among them, the method of image enhancement processing is set according to needs. For example, image enhancement can include methods such as noise removal, contrast enhancement, and color enhancement. Noise removal is a processing process of transforming each pixel in the input image using a Fourier transform algorithm. Contrast enhancement is to enhance the contrast between black and white to make the boundary or edge in the image more obvious. Color enhancement is to convert the image of each pixel in the image to a color space of another spatial domain, subtract the minimum value of all pixels in the spatial domain from the pixel value of each pixel, and then convert it into the space of the original domain. Grayscale processing is a method of calculating the pixel value of each pixel in the input image using a grayscale algorithm to obtain an image.
[0042] Among them, the training process of the target detection network inputs the trained sample image containing the owner area label, and the owner features of the sample image are extracted through the network containing parameters. The parameters of the network can be obtained through transfer learning, and classification is performed according to the extracted owner features to classify the sample image containing the owner area range. The sample recognition result, the loss value between the sample recognition result and the owner area label associated with the sample image is calculated, and the parameters of the target detection network are iteratively updated according to the loss value until the loss value reaches the convergence condition, wherein the convergence condition can be that the loss value no longer changes or the number of iterations reaches a preset number, and the training is stopped, and the target detection network after convergence is recorded as the trained target detection network.
[0043] In one embodiment, the vehicle sub-attribute recognition is performed on the captured vehicle image to obtain a second recognition result corresponding to the original data, including:
[0044] Perform license plate recognition on the captured vehicle image to obtain the license plate number of the captured vehicle image.
[0045] It can be understood that license plate recognition is the process of identifying the license plate number of a vehicle in an input vehicle snapshot image. The process of license plate recognition is as follows: first, the license plate is located in the vehicle snapshot image. Since the color combinations of the characters on the license plate and the background behind the characters generally include blue background with white characters, yellow background with black characters, white background with red characters, green background with white characters, and black background with white characters, different color channels (RGB channels, i.e., including R channel, G channel and B channel) can be used to clearly distinguish the area from the background. For example, for a license plate with white characters on a blue background, when the blue channel (B channel) is separated from the vehicle snapshot image, the license plate area is a bright rectangle, and the characters in the license plate are not presented in the rectangular area, because the blue (255, 0, 0) and white (255, 255, 255) in the color channel are indistinguishable in the blue channel, and the same is true for the white background. The red channel (R channel) can be used for license plates with black characters, and the green channel (G channel) can be used for license plates with white characters on a green background to clearly show the location of the license plate area. The edge of the area is then extracted, and the boundary part of the area where the local brightness changes significantly is detected. The edge of the license plate can be detected, and the area surrounded by the edge of the license plate can be located. Secondly, the image corresponding to the located license plate area is processed by image enhancement technology, such as noise removal method and contrast enhancement method, so that the text outline in the license plate area can be enhanced, and the image corresponding to the license plate area after image enhancement technology is segmented to obtain separate character images. Finally, character normalization and character recognition are performed on each separate character image, and the recognized characters are merged into the license plate number.
[0046] Among them, character normalization processing is to normalize the character image with marked serial numbers from 0 to 1, and character recognition can be a recognition process of first binarizing the character image and scaling its size to the size of the template in the character database, then matching it with all templates, and finally selecting the best match as the result. It can also use a character recognition network to extract features of the character image, and then use the obtained features to classify the trained classifier to obtain the recognition process with the highest probability after classification. The training process of the character recognition network is to input a trained character image containing character labels, extract character features of the character image through the network containing parameters, the parameters of the network can be obtained through transfer learning, and classification is performed according to the extracted character features to classify the character recognition results of the corresponding characters, calculate the loss value between the character recognition result and the character label associated with the character image, and iteratively update the parameters of the character recognition network according to the loss value until the loss value reaches the convergence condition, wherein the convergence condition can be that the loss value no longer changes or the number of iterations reaches a preset number, and the training is stopped, and the converged character recognition network is recorded as the trained character recognition network.
[0047] The vehicle model attributes are recognized on the captured vehicle image to obtain the attribute results of the captured vehicle image.
[0048] It can be understood that vehicle type attribute recognition is the process of identifying the vehicle type in the vehicle captured image. The process of vehicle type attribute recognition is: extracting vehicle type features from the input vehicle captured image through the trained vehicle type detection model, classifying it according to the extracted vehicle type features, classifying the probability values of various vehicle types, and obtaining the vehicle type category corresponding to the maximum probability value as the attribute result. The attribute result reflects the vehicle type category of the vehicle in the vehicle captured image. The vehicle type feature is a feature related to the vehicle type category. Among them, the vehicle type categories include taxis, buses, private cars, trucks, engineering vehicles, etc. The training process of the vehicle type detection network The vehicle image with the vehicle model label is input for training, and the vehicle model features are extracted from the vehicle model image through the network with parameters. The parameters of the network can be obtained through transfer learning, and classification is performed according to the extracted vehicle model features to classify the vehicle model recognition results of the corresponding vehicles, and the loss value between the vehicle model recognition result and the vehicle model label associated with the vehicle model image is calculated. The parameters of the vehicle model detection network are iteratively updated according to the loss value until the loss value reaches the convergence condition. The convergence condition can be that the loss value no longer changes or the number of iterations reaches a preset number, and the training is stopped, and the vehicle model detection network after convergence is recorded as the trained vehicle model detection network.
[0049] The license plate number and attribute result of the captured vehicle image are recorded as the second recognition result.
[0050] Understandably, the license plate number and attribute result of the captured vehicle image are determined as the second recognition result, and the second recognition result includes the vehicle owner area image, the license plate number and the attribute result.
[0051] The present invention realizes obtaining the license plate number of the vehicle captured image by performing license plate recognition on the vehicle captured image; obtaining the attribute result of the vehicle captured image by performing vehicle model attribute recognition on the vehicle captured image; and recording the license plate number and the attribute result of the vehicle captured image as the second recognition result. In this way, the license plate recognition and vehicle model attribute recognition can be used to automatically recognize the license plate number and the attribute result including the vehicle model category, and finally determine the second recognition result of the vehicle sub-attribute recognition, without manual recognition, reducing labor costs, and improving the efficiency and accuracy of vehicle sub-attribute recognition.
[0052] S105, recording the first recognition result and the second recognition result corresponding to the original data as image data.
[0053] It can be understood that the image data corresponding to one original data includes the first recognition result and the second recognition result, so that after traversing all the original data, the image data corresponding to each original data is obtained.
[0054] The present invention achieves the following steps: acquiring a plurality of original data; for each original data, performing vehicle recognition on a captured image in the original data to obtain a first recognition result corresponding to the original data; based on the first recognition result corresponding to each original data, performing vehicle screening on all original data to screen out original data containing vehicles; based on the vehicle area in the first recognition result, extracting the captured image in the screened original data to obtain at least one captured image of a vehicle, and performing vehicle sub-attribute recognition on the captured image of the vehicle to obtain a second recognition result corresponding to the original data; recording the first recognition result and the second recognition result corresponding to the original data as image data, so that it is possible to automatically identify whether the original data contains a vehicle through vehicle recognition, and automatically screen out the original data containing the vehicle, and perform vehicle sub-attribute recognition on the screened original data to automatically identify the second recognition result of the focused attribute of the vehicle, thereby determining the image data, and therefore, it is possible to automatically interpret the focused attribute of the vehicle in the original data, and interpret the large-capacity original data into small-capacity information data of each sub-attribute, thereby greatly reducing the data processing operations for subsequent processing, compressing the capacity, and improving the efficiency and diversity of image data output.
[0055] S20, determining the number of target vehicles of a specific vehicle type that appear in the area to be monitored within a preset time period and the owner information of the target vehicles based on the plurality of image data.
[0056] It can be understood that the specific vehicle type is a vehicle type or vehicle attribute category with pre-set abnormal attribute rules, for example: the preset abnormal attribute rule is the license plate number of a certain area, such as license plate numbers starting with Guangdong A or Liao C, or the preset abnormal attribute rule is a certain vehicle type category, such as taxis, buses, etc. The process of determining the target vehicle of the specific vehicle type that appears in the monitored area within a preset time period is to filter out image data that meets the specific vehicle type from all image data, and determine the vehicle in the filtered image data as the target vehicle, and then count the number of target vehicles in the filtered image data, and obtain the owner information associated with the license plate number of the target vehicle through the license plate number of the target vehicle.
[0057] In one embodiment, in step S20, determining the number of target vehicles of a specific vehicle type and the owner information of the target vehicles that appear in the monitored area within a preset time period based on the plurality of image data includes:
[0058] A preset number of continuous historical time periods adjacent to the preset time period is obtained; each historical time period is a time period with the same period as the preset time period.
[0059] Understandably, the preset number can be set according to demand, for example, the preset number can be 5 or 7, the historical time period can be the same time period as the preset time period, or it can be a time period of a complete day in the historical record, for example: obtaining a preset number of consecutive historical time periods adjacent to the preset time period can be: the preset time period is the time period from the current time (from 14:00 to 14:59), the preset number is 7, and the adjacent preset number of consecutive historical time periods are the 7 days before the current day (14:00 to 14:59).
[0060] The historical snapshot values associated with all historical time periods are averaged to obtain the historical average.
[0061] It can be understood that each historical time period is associated with a historical snapshot value, which is the number of vehicles captured in the historical time period. For a historical segment, the historical snapshot values associated with each historical time period are obtained. The mean processing is to take the average value of the obtained historical snapshot values. After the mean processing, the historical mean corresponding to the preset time period is obtained.
[0062] The statistics of the historical snapshot values are the statistics of the total number of vehicle regions of the first recognition results in the image data corresponding to the monitoring time falling into the same historical time period.
[0063] All image data are captured and counted to obtain the captured values to be compared.
[0064] It can be understood that the snapshot statistics are statistics of the total number of vehicle areas of the first recognition result in the image data corresponding to the monitoring time falling into the same preset time period, and the statistical results are recorded as the snapshot values to be compared.
[0065] Determine whether the snapshot value to be compared is greater than the historical average.
[0066] When the captured value to be compared is greater than the historical average, all image data are screened for specific vehicle types to filter out target image data corresponding to the target vehicle.
[0067] It can be understood that if the snapshot value to be compared is greater than the historical average, the screening of specific vehicle types is a screening process for screening out image data corresponding to the target vehicle that meets the specific vehicle type. The target vehicle that meets the specific vehicle type can be determined through the screened image data, and the image data corresponding to the target vehicle is recorded as target image data.
[0068] In one embodiment, all image data are screened for specific vehicle types to screen out target image data corresponding to the target vehicle, including:
[0069] Determine the exceeded value within the preset time period based on the snapshot value to be compared and the historical average.
[0070] It can be understood that the size of the historical snapshot value and the snapshot value to be compared corresponding to the preset time period is judged. If the historical snapshot value is greater than or equal to the snapshot value to be compared, the excess value within the preset time period is determined to be zero. If the historical snapshot value is less than the snapshot value to be compared, the difference between the snapshot value to be compared and the historical snapshot value is calculated, and the ratio of the difference to the historical snapshot value is determined as the excess value within the preset time period.
[0071] If the excess value corresponding to the preset time period is greater than or equal to the preset threshold, the image data corresponding to the excess value is screened for a specific vehicle type to screen out target image data corresponding to the target vehicle.
[0072] It is understandable that if the detected exceeded value is greater than or equal to the preset threshold, the preset threshold can be set according to demand, for example, the preset threshold is set to 20%, indicating that the snapshot value to be compared exceeds the floating standard of the historical snapshot value and there is aggregation.
[0073] If the excess value corresponding to the preset time period is less than the preset threshold, the image data corresponding to the excess value will not be screened for specific vehicle types. It can be understood that if the detected excess value is less than the preset threshold, it means that there is no aggregation.
[0074] In this way, it is possible to determine through the excess value, and then through the judgment between the excess value and the preset threshold, automatically output the image data with abnormal aggregation, and then screen for specific abnormal vehicle types, and automatically screen out the target image data of the target vehicle, thereby realizing automatic screening of which image data contains the target vehicle without manual judgment.
[0075] In one embodiment, after determining whether the snapshot value to be compared is greater than the historical average, the following steps are performed:
[0076] When the snapshot value to be compared is less than or equal to the historical average, the abnormal aggregation detection result of the absence of abnormal aggregation in the monitored area within the preset time period is determined. It can be understood that if the snapshot value to be compared is less than or equal to the historical average, it means that there is no aggregation in the monitored area.
[0077] The number of all target image data is counted to obtain the number of vehicles, and the owner information of the license plate number in each target image data is obtained to obtain the owner information of the target vehicle.
[0078] It can be understood that quantity statistics is the process of counting the number of target vehicles in all target image data. During the statistical process, the same target vehicles will be deduplicated, so that the number of vehicles obtained is the total number of different target vehicles. The deduplication of the same target vehicle can be achieved by deduplicating the same license plate number. The license plate numbers of the target vehicles in all target image data are deduplicated to obtain the license plate numbers to be checked. The owner information associated with the license plate numbers to be checked is searched from the owner information database, and the found owner information is associated with the corresponding target vehicle to obtain the owner information of each target vehicle.
[0079] Among them, the owner information database stores the association between all license plate numbers and owner information. One license plate number can be associated with multiple owner information. Since taxis have owners who work in both day and night shifts, the owner associated with the query can be determined by matching the preset time period with the time range of the two shifts.
[0080] The present invention achieves this by acquiring a preset number of continuous historical time periods adjacent to a preset time period; performing mean processing on historical snapshot values associated with all historical time periods to obtain a historical mean; performing snapshot statistics on all image data to obtain a snapshot value to be compared; judging whether the snapshot value to be compared is greater than the historical mean; when the snapshot value to be compared is greater than the historical mean, screening all image data for a specific vehicle type to screen out target image data corresponding to the target vehicle; performing quantitative statistics on all target image data to obtain the number of vehicles, and obtaining the owner information of the license plate number in each target image data to obtain the owner information of the target vehicle. In this way, the historical mean of the historical time periods of the same dimension can be automatically compared, and it can be judged whether the current image data has abnormal aggregation through the historical mean. In the case of abnormal aggregation, the vehicle number and owner information of the target vehicle are automatically counted without manual judgment and confirmation, thereby achieving scientific and objective screening of the vehicle number and owner information under abnormal aggregation.
[0081] S30, based on the number of vehicles, the owner information of the target vehicle and a preset abnormal aggregation blacklist, determining the proportion of owners with abnormal aggregation records.
[0082] Understandably, the abnormal gathering blacklist is a list of persons who have a history of abnormal gathering, potential risk of abnormal gathering, or have participated in abnormal gathering.
[0083] In one embodiment, in step S30, based on the number of vehicles, the owner information of the target vehicle, and a preset abnormal aggregation blacklist, determining the percentage of owners with abnormal aggregation records includes:
[0084] The owner information of the target vehicle is matched with the abnormal aggregation blacklist to obtain the target vehicle with abnormal aggregation records.
[0085] It can be understood that the text matching algorithm is used to match the owner information of the target vehicle with each name in the abnormal aggregation blacklist, and the target vehicle corresponding to the owner information that matches any name in the abnormal aggregation blacklist is recorded as the target vehicle with an abnormal aggregation record. The matching process is to use the text matching algorithm to calculate the degree of complete matching between the owner information and each name in the abnormal aggregation blacklist, and determine the owner information and name that completely match as a match between the owner information and the abnormal aggregation blacklist.
[0086] The text matching algorithm can be set according to the requirements, for example, the text matching algorithm is BF, RK, KMP, BM, Sunday and other algorithms. Preferably, the text matching algorithm is KMP, also known as the fast string matching algorithm.
[0087] In one embodiment, the vehicle owner information includes a vehicle owner area image; the vehicle owner information of the target vehicle is matched with the abnormal aggregation blacklist to obtain the target vehicle with abnormal aggregation records, including:
[0088] The vehicle owner area image in the vehicle owner information is input into the blacklist face detection model, and the facial features of the vehicle owner area image are extracted through the blacklist face detection model.
[0089] It can be understood that the blacklist face detection model is a neural network model that is trained and completed by using face images included in the abnormal aggregation blacklist. The blacklist face detection model can identify whether the input image containing a face is a model of a face included in the abnormal aggregation blacklist. Each owner area image is input into the blacklist face detection model, and the blacklist face detection model is used to extract facial features from the owner area image to obtain a facial feature vector to be compared. The extraction process is to first grayscale the two-dimensional three-channel owner area image, convert the owner area image into an image of one channel, and then convolve the channel image, convolve through different convolution kernels, extract facial features, and finally perform dimensionality reduction processing on the extracted feature vector to obtain a one-dimensional vector. For example: the facial features are converted end to end into a column vector. Assuming the image size is 20*20, then this vector is 400-dimensional.
[0090] Among them, the abnormal gathering blacklist is a list of people with a historical record of abnormal gathering, potential risks of abnormal gathering, or participation in abnormal gatherings, such as a previous record of abnormal gatherings such as fighting, etc., and the facial features are the features of the faces of people on the abnormal gathering blacklist.
[0091] The extracted facial features are compared with the facial features of people included in the preset abnormal aggregation blacklist to determine the comparison result.
[0092] It can be understood that a similarity comparison is performed based on the extracted facial features (facial feature map to be compared) and the facial features of the persons included in the abnormal aggregation blacklist (facial feature vectors of the persons). The similarity comparison is a difference comparison of the facial features between the two facial feature vectors. The cosine similarity algorithm can be used to calculate the similarity between the two facial feature vectors, and the corresponding probability value is mapped according to the similarity, so that the probability value corresponding to each person can be obtained, and the maximum probability value among the probability values corresponding to all persons is obtained, which is determined as the comparison result.
[0093] The cosine similarity algorithm is an algorithm for calculating the cosine value between two vector values as the similarity.
[0094] Based on the comparison results, target vehicles with abnormal aggregation records are obtained.
[0095] It can be understood that if the comparison result is greater than or equal to the preset probability threshold, it is determined that the target vehicle corresponding to the owner area corresponding to the comparison result is a target vehicle with an abnormal aggregation record. If the comparison result is less than the preset probability threshold, it is determined that the target vehicle corresponding to the owner area corresponding to the comparison result is not a target vehicle with an abnormal aggregation record.
[0096] The present invention realizes that the facial features of the vehicle owner area image in the vehicle owner information are extracted by inputting the vehicle owner area image into a blacklist face detection model, and the facial features of the vehicle owner area image are extracted by the blacklist face detection model; the extracted facial features are compared with the facial features of persons included in a preset abnormal aggregation blacklist to determine the comparison result; based on the comparison result, a target vehicle with an abnormal aggregation record is obtained, thus, the facial features of the vehicle owner in the vehicle owner area image are extracted by the blacklist face detection model, and the extracted facial features are automatically compared with the facial features of persons in the abnormal aggregation blacklist to obtain the comparison result, and the target vehicle with the abnormal aggregation record is automatically determined, thereby improving the recognition efficiency and accuracy of the target vehicle with the abnormal aggregation record.
[0097] Based on the ratio of the total number of target vehicles with abnormal aggregation records to the number of vehicles, the owner ratio value is obtained.
[0098] Understandably, the ratio of the total number of target vehicles with abnormal aggregation records to the number of vehicles is calculated, and the ratio is recorded as the vehicle owner ratio value.
[0099] The present invention achieves that the target vehicles with abnormal aggregation records are obtained by matching the owner information of the target vehicles with the abnormal aggregation blacklist; the owner ratio value is obtained based on the ratio of the total number of target vehicles with abnormal aggregation records to the number of vehicles. In this way, the target vehicles with abnormal aggregation records can be automatically matched through the abnormal aggregation blacklist to obtain the owner ratio value, and the ratio of the target vehicles with abnormal aggregation records in the number of vehicles can be automatically output, thereby providing a data basis for subsequent abnormal aggregation detection.
[0100] S40, determining the abnormal aggregation detection result of the monitored area within a preset time period according to the number of vehicles and the proportion of vehicle owners.
[0101] Understandably, the process of determining the abnormal aggregation detection result can be as follows: recording the difference between the number of vehicles and the preset number of aggregations as the aggregation number index, obtaining the abnormal aggregation level value corresponding to the aggregation number index from the aggregation number index-abnormal aggregation level value comparison table, and mapping the abnormal aggregation level value corresponding to the car owner proportion value from the car owner proportion value-abnormal aggregation level value mapping table according to the car owner proportion value, and recording the largest abnormal aggregation level value or the mean / sum of the abnormal aggregation level value corresponding to the aggregation number index and the abnormal aggregation level value corresponding to the car owner proportion value as the abnormal aggregation detection result. The abnormal aggregation detection result reflects the level value of the abnormal aggregation degree reflected in the monitored area within the preset time period. The larger the level value, the higher the degree of abnormal aggregation. According to the level range that different level values fall into, the corresponding high, medium and low color warnings are determined to facilitate the execution of further related matters, such as police dispatch or patrolling, etc.
[0102] Among them, the process of determining the abnormal aggregation detection result can also be: multiplying the number of vehicles by the proportion of car owners to obtain the number of blacklisted vehicles. If the number of blacklisted vehicles is greater than the preset number, subtract the number of blacklisted vehicles from the preset number to obtain the warning value. Based on the warning value, the level value of the abnormal aggregation detection result is determined, that is, the corresponding level value of the abnormal aggregation degree is mapped according to the range in which the warning value falls. If the number of blacklisted vehicles is less than or equal to the preset number, the warning value is zero, and the abnormal aggregation detection result is determined to be risk-free.
[0103] Among them, the process of determining the abnormal aggregation detection result also includes: taking the weighted sum of the number of vehicles and the proportion of car owners to obtain the predicted number of blacklists. The weights of the number of vehicles and the proportion of car owners can be set according to needs. The sum of the weight of the number of vehicles and the weight of the proportion of car owners is one. For example, the weight of the number of vehicles is 0.4, and the weight of the proportion of car owners is 0.6. If the predicted number of blacklists is greater than the preset number, the predicted number of blacklists is subtracted from the preset number to obtain a warning value. Based on the warning value, the level value of the abnormal aggregation detection result is determined, that is, the corresponding level value of the abnormal aggregation degree is mapped according to the range in which the warning value falls. If the predicted number of blacklists is less than or equal to the preset number, the warning value is zero, and the abnormal aggregation detection result is determined to be risk-free.
[0104] The present invention realizes the acquisition of multiple image data obtained by shooting in the area to be monitored within a preset time period; determining the number of target vehicles of a specific vehicle type appearing in the area to be monitored within the preset time period and the owner information of the target vehicles based on the multiple image data; determining the proportion of owners with abnormal aggregation records based on the number of vehicles, the owner information of the target vehicles and a preset abnormal aggregation blacklist; determining the abnormal aggregation detection result of the area to be monitored within the preset time period based on the number of vehicles and the proportion of owners. Therefore, the present invention obtains the image data shot in the area to be monitored within the preset time period, and automatically determines the specific vehicle type from all the image data The number of target vehicles and information of vehicle owners are taken into account, and based on the abnormal aggregation blacklist, the proportion of vehicle owners is automatically determined. The number of vehicles and the proportion of vehicle owners are comprehensively considered, and the abnormal aggregation detection result of the area to be monitored is automatically output. In this way, image data can be obtained from a large amount of captured data, and the number of target vehicles and information of vehicle owners of a specific vehicle type can be determined. Based on the abnormal aggregation blacklist, it is possible to objectively, accurately and automatically identify whether there is abnormal aggregation in the abnormal aggregation detection result of the area to be monitored. There is no need to manually identify whether there is aggregation in the image data, which greatly saves labor costs and energy, and improves the accuracy and correctness of abnormal aggregation monitoring of vehicles.
[0105] In one embodiment, in step S40, determining the abnormal aggregation detection result of the monitored area within a preset time period according to the number of vehicles and the proportion of vehicle owners includes:
[0106] The aggregation results are determined based on the number of vehicles and the proportion of vehicle owners.
[0107] Understandably, the process of determining the clustering result can be: weighted sum of the number of vehicles and the proportion of car owners to obtain the clustering result. The weights of the number of vehicles and the proportion of car owners can be set according to needs. The weights of the number of vehicles and the proportion of car owners are not greater than 1, and the sum of the weights of the number of vehicles and the proportion of car owners is not greater than 0.9. For example, the weight of the number of vehicles is 0.4, and the weight of the proportion of car owners is 0.3. The process of determining the clustering result can also be: multiplying the number of vehicles by the proportion of car owners to obtain the clustering result.
[0108] The risk level associated with the area to be monitored is obtained, and the abnormal aggregation detection result is determined based on the aggregation result and the risk level.
[0109] It can be understood that a monitored area is associated with a risk level, which reflects the degree of risk of aggregation in the monitored area. The risk level will be continuously updated over time. For example, the risk level is determined by accumulating the number of abnormal aggregation detection results in which abnormal aggregation occurs. That is, the risk level of the monitored area will be analyzed along with the relevant historical data of the monitored area to obtain the risk level associated with the monitored area. According to the aggregation result and the risk level, the process of determining the abnormal aggregation detection result can be: weighted summing the aggregation result and the risk level to obtain the abnormal aggregation detection result, wherein the weight of the aggregation result is one, and the sum of the weight of the risk level, the weight of the number of vehicles and the weight of the proportion of vehicle owners is one.
[0110] The present invention achieves the goal of determining the aggregation result based on the number of vehicles and the proportion of vehicle owners; obtaining the risk level associated with the area to be monitored, and determining the abnormal aggregation detection result based on the aggregation result and the risk level. In this way, the risk level of the area to be monitored can be introduced to comprehensively output the abnormal aggregation detection result, which more accurately and objectively reflects the result of abnormal aggregation whether there is any in the area to be monitored within the current preset time period, thereby improving the accuracy of abnormal vehicle aggregation monitoring.
[0111] For each aggregated data, the risk level value, traffic value and blacklist ratio value corresponding to the aggregated data are multiplied and summed with their respective preset weights, where the sum of all preset weights is 1, thereby obtaining the abnormal aggregation value of the aggregated data. For example, the weight corresponding to the risk level value is 0.5, the weight corresponding to the traffic value is 0.2, and the weight corresponding to the blacklist ratio value is 0.3. The abnormal warning is to map the corresponding level according to the abnormal aggregation value. For example, the value in the range of 0.1 to 0.5 corresponds to the low level, the value in the range of 0.5 to 0.9 corresponds to the medium level, and so on. All the levels of abnormal aggregation are classified and summarized to obtain the abnormal vehicle aggregation monitoring result. The corresponding monitoring area identification can be quickly displayed on the visual interface through the abnormal vehicle aggregation monitoring result, or different colors can be displayed on the visual interface through different levels, so as to perform further monitoring or execute related transactions.
[0112] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.
[0113] In one embodiment, a vehicle abnormal gathering monitoring device is provided, and the vehicle abnormal gathering monitoring device corresponds one-to-one with the vehicle abnormal gathering monitoring method in the above embodiment. Figure 4 As shown, the vehicle abnormal gathering monitoring device includes an acquisition module 11, a statistical module 12, a determination module 13 and an output module 14. The functional modules are described in detail as follows:
[0114] An acquisition module 11 is used to acquire a plurality of image data captured in the monitored area within a preset time period;
[0115] A statistical module 12, used to determine the number of target vehicles of a specific vehicle type appearing in the monitored area within a preset time period and the owner information of the target vehicles according to the plurality of image data;
[0116] A determination module 13, configured to determine a percentage of vehicle owners having abnormal aggregation records based on the number of vehicles, the owner information of the target vehicle, and a preset abnormal aggregation blacklist;
[0117] The output module 14 is used to determine the abnormal aggregation detection result of the monitored area within a preset time period according to the number of vehicles and the proportion of vehicle owners.
[0118] For the specific definition of the abnormal vehicle aggregation monitoring device, please refer to the definition of the abnormal vehicle aggregation monitoring method mentioned above, which will not be repeated here. Each module in the above-mentioned abnormal vehicle aggregation monitoring device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0119] In one embodiment, a computer device is provided. The computer device may be a client or a server. The internal structure diagram thereof may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a readable storage medium and an internal memory. The readable storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the readable storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for monitoring abnormal aggregation of vehicles is implemented.
[0120] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for monitoring abnormal aggregation of vehicles in the above-mentioned embodiment is implemented.
[0121] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the vehicle abnormal aggregation monitoring method in the above embodiment is implemented.
[0122] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0123] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0124] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.
Claims
1. A method for monitoring abnormal vehicle aggregation, characterized in that: include: Acquire multiple image data captured in the area to be monitored within a preset time period; Determine the number of target vehicles of a specific vehicle type that appear in the monitored area within a preset time period and the owner information of the target vehicles based on the plurality of image data; Based on the number of vehicles, the owner information of the target vehicle and a preset abnormal aggregation blacklist, determining a percentage of owners with abnormal aggregation records; According to the number of vehicles and the proportion of vehicle owners, the abnormal aggregation detection result of the monitored area within a preset time period is determined.
2. The method for monitoring abnormal vehicle aggregation according to claim 1, characterized in that: The determining, according to the number of vehicles and the proportion of vehicle owners, the abnormal aggregation detection result of the monitored area within a preset time period includes: Determine a clustering result according to the number of vehicles and the proportion of vehicle owners; A risk level associated with the area to be monitored is obtained, and the abnormal aggregation detection result is determined according to the aggregation result and the risk level.
3. The method for monitoring abnormal vehicle aggregation according to claim 1, characterized in that: The step of acquiring a plurality of image data captured in the monitored area within a preset time period includes: Get multiple raw data; For each of the original data, vehicle recognition is performed on the captured image in the original data to obtain a first recognition result corresponding to the original data; the first recognition result reflects whether the captured image contains a vehicle; Based on the first recognition result corresponding to each of the original data, screening all the original data for vehicles, and screening out the original data containing vehicles; Based on the vehicle area in the first recognition result, extract the captured image in the filtered original data to obtain at least one vehicle captured image, and perform vehicle sub-attribute recognition on the vehicle captured image to obtain a second recognition result corresponding to the original data; The first recognition result and the second recognition result corresponding to the original data are recorded as the image data.
4. The method for monitoring abnormal vehicle aggregation as claimed in claim 3, characterized in that: The performing vehicle sub-attribute recognition on the captured vehicle image to obtain a second recognition result corresponding to the original data includes: Performing license plate recognition on the captured vehicle image to obtain the license plate number of the captured vehicle image; Performing vehicle type attribute recognition on the captured vehicle image to obtain an attribute result of the captured vehicle image; The license plate number and the attribute result of the captured vehicle image are recorded as the second recognition result.
5. The method for monitoring abnormal vehicle aggregation according to claim 4, characterized in that: The method of determining the number of target vehicles of a specific vehicle type that appear in the monitored area within a preset time period and the owner information of the target vehicles based on the plurality of image data includes: Acquire a preset number of continuous historical time periods adjacent to the preset time period; each of the historical time periods is a time period having the same period as the preset time period; Performing average processing on all historical snapshot values associated with the historical time period to obtain a historical average; Snapshot statistics are performed on all the image data to obtain the snapshot values to be compared; Determine whether the captured value to be compared is greater than the historical average value; When the captured value to be compared is greater than the historical average value, all the image data are screened for a specific vehicle type to screen out target image data corresponding to the target vehicle; Counting the number of all the target image data to obtain the number of vehicles, and acquiring the owner information of the license plate number in each of the target image data to obtain the owner information of the target vehicle.
6. The method for monitoring abnormal vehicle aggregation according to claim 5, characterized in that: The determining, based on the number of vehicles, the owner information of the target vehicle, and a preset abnormal aggregation blacklist, a percentage of owners having abnormal aggregation records includes: Matching the owner information of the target vehicle with the abnormal aggregation blacklist to obtain the target vehicle with abnormal aggregation records; The vehicle owner ratio value is obtained based on the ratio of the total number of target vehicles with abnormal aggregation records to the number of vehicles.
7. The method for monitoring abnormal vehicle aggregation according to claim 6, characterized in that: The vehicle owner information includes a vehicle owner area image; the matching of the vehicle owner information of the target vehicle with the abnormal aggregation blacklist to obtain the target vehicle with abnormal aggregation records includes: Inputting the vehicle owner area image in the vehicle owner information into a blacklist face detection model, and extracting facial features from the vehicle owner area image through the blacklist face detection model; Compare the extracted facial features with the facial features of persons included in the preset abnormal aggregation blacklist to determine the comparison result; Based on the comparison results, target vehicles with abnormal aggregation records are obtained.
8. A vehicle abnormal gathering monitoring device, characterized in that: include: An acquisition module is used to acquire a plurality of image data captured in the monitored area within a preset time period; A statistical module, used to determine the number of target vehicles of a specific vehicle type appearing in the monitored area within a preset time period and the owner information of the target vehicles according to the plurality of image data; A determination module, configured to determine a percentage of vehicle owners having abnormal aggregation records based on the number of vehicles, the owner information of the target vehicle, and a preset abnormal aggregation blacklist; The output module is used to determine the abnormal aggregation detection result of the monitored area within a preset time period according to the number of vehicles and the proportion of vehicle owners.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the vehicle abnormal aggregation monitoring method as described in any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the abnormal vehicle aggregation monitoring method as described in any one of claims 1 to 7 is implemented.
