Vehicle monitoring methods, devices, electronic equipment and storage media
By using deep learning models and behavior recognition technology to compare features of vehicle images, the problem of low efficiency in manual review in existing vehicle supervision is solved, enabling rapid identification and management of suspicious vehicles and improving the supervision efficiency of vehicle snatching cases.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN INTELLIFUSION TECHNOLOGIES CO LTD
- Filing Date
- 2023-11-16
- Publication Date
- 2026-07-17
Smart Images
Figure CN117576642B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of smart cities, and more particularly to a vehicle monitoring method, device, electronic device, and storage medium. Background Technology
[0002] Vehicle snatching is a common crime in many cities, posing a significant threat to society. Current surveillance technologies often rely on manual processing to extract the characteristics and location information of vehicle snatching cases. This not only increases workload but also increases the risk of misjudgments during manual review. Furthermore, manual review becomes even more burdensome when there are many cases. Therefore, existing vehicle surveillance methods suffer from the problems of high workload, inaccurate judgments, and low efficiency. Summary of the Invention
[0003] This invention provides a vehicle monitoring method aimed at addressing the problems of existing vehicle monitoring methods, such as high workload for manual review, inaccurate judgment, and low efficiency. Based on the extraction of attribute features from historical vehicle robbery cases, a deep learning model can compare the time, location, and behavioral attribute features of the vehicle image to be identified with the attribute features of vehicles in vehicle robbery cases to determine whether the vehicle to be identified is suspicious. Furthermore, by comparing the attribute features of suspicious vehicles with those of vehicles involved in the cases, it can further determine whether the suspicious vehicle was the one used in the vehicle robbery. This allows for rapid identification of suspicious vehicles, speeds up response times, and improves the efficiency of monitoring vehicles involved in vehicle robberies.
[0004] In a first aspect, embodiments of the present invention provide a vehicle monitoring method, characterized in that the method includes the following steps:
[0005] Acquire an image of the vehicle to be identified;
[0006] Based on the vehicle image to be identified and the vehicle robbery case attribute features, a first suspicious vehicle image is determined from the vehicle image to be identified, and the vehicle robbery case attribute features are extracted from historical vehicle robbery case information;
[0007] Based on the location information of the vehicle robbery case, a second suspicious vehicle image is determined from the first suspicious vehicle image, wherein the location information of the vehicle robbery case is determined by the historical vehicle robbery case information;
[0008] The second suspicious vehicle image is processed by a preset behavior recognition model to identify the vehicle behavior, and a third suspicious vehicle image is determined from the second suspicious vehicle image.
[0009] The third suspicious vehicle image is compared with the vehicle image in the criminal record to determine whether the vehicle in the third suspicious vehicle image is the vehicle used to commit the vehicle robbery, and the vehicle used to commit the vehicle robbery is then managed.
[0010] Optionally, the vehicle monitoring method is characterized in that, based on the image of the vehicle to be identified and the attribute features of the vehicle robbery case, determining the first suspicious vehicle image from the image of the vehicle to be identified includes:
[0011] The image of the vehicle to be identified is identified to obtain images of each target vehicle in the image of the vehicle to be identified, and the image of each target vehicle includes at least one target vehicle.
[0012] The structured engine is used to extract features from the images of each target vehicle to obtain the attribute features of each target vehicle.
[0013] The attribute features of each target vehicle are compared with the attribute features of the vehicle robbery case; if the attribute features of the target vehicle are successfully matched with the attribute features of the vehicle robbery case, the image of the target vehicle that is successfully matched is identified as the first suspicious vehicle image.
[0014] Optionally, the vehicle monitoring method is characterized in that, before determining the second suspicious vehicle image based on the location information of the vehicle robbery case and the first suspicious vehicle image, it includes:
[0015] Based on the location information of the vehicle robbery case, determine the locations near the case location;
[0016] Based on the latitude and longitude of the aforementioned attachment points, determine the area where the points are concentrated;
[0017] The area where the points are concentrated is taken as the target area.
[0018] Optionally, the vehicle monitoring method is characterized in that determining the second suspicious vehicle image based on the location information of the vehicle robbery case and the first suspicious vehicle image includes:
[0019] If the first suspicious vehicle image contains a license plate, then the first count detection is performed in the target area using the license plate of the vehicle in the first suspicious vehicle image to obtain the number of times the vehicle in the first suspicious vehicle image appears in the target area;
[0020] If the first suspicious vehicle image does not have a license plate, a second number detection is performed in the target area based on the vehicle features of the vehicle in the first suspicious vehicle image to obtain the number of times the vehicle in the first suspicious vehicle image appears in the target area;
[0021] Based on the number of times the vehicle appears in the target area in the first suspicious vehicle image, a second suspicious vehicle image is determined from the first suspicious vehicle image.
[0022] Optionally, the vehicle monitoring method is characterized in that, the step of performing vehicle behavior recognition processing on the second suspicious vehicle image using a preset behavior recognition model to determine the third suspicious vehicle image includes:
[0023] A preset clustering time is obtained, which can be adjusted according to the size of the target region;
[0024] Based on the preset clustering time, a sliding window algorithm is used to count the vehicles in the second suspicious vehicle images within the target area to obtain the dwell data of the vehicles in the second suspicious vehicle images. The dwell data includes the number of times the vehicles in the second suspicious vehicle images were photographed in the target area and the vehicle images photographed during the dwell time.
[0025] Based on a preset dwell threshold and the number of dwell times, a third suspicious vehicle image is determined from the second suspicious vehicle image.
[0026] Optionally, the vehicle monitoring method is characterized in that, the step of comparing the features of the third suspicious vehicle image with the vehicle image in the criminal record to determine whether the vehicle in the third suspicious vehicle image is the vehicle used to commit the vehicle robbery includes:
[0027] The license plate of the vehicle in the third suspicious vehicle image is compared with the license plate of the vehicle with the criminal record to obtain the first comparison result;
[0028] The vehicle features in the third suspicious vehicle image are compared with the vehicle features of the vehicle with the criminal record to obtain a second comparison result;
[0029] Based on at least one of the first comparison result and the second comparison result, determine whether the vehicle in the third suspicious vehicle image is the vehicle used to commit vehicle robbery.
[0030] Optionally, the vehicle monitoring method is characterized in that the management of the vehicle involved in the vehicle hijacking further includes:
[0031] If it is determined that the vehicle in the third suspicious vehicle image is the vehicle used to commit the vehicle robbery, then the vehicle characteristics in the third suspicious vehicle image are entered into the database and the case is recorded so that the vehicle used to commit the vehicle robbery can be tracked subsequently.
[0032] If it is determined that the third suspicious vehicle is not the vehicle that committed the vehicle robbery, then the vehicle characteristics of the vehicle that committed the vehicle robbery are taken as failure attributes, so that the method for extracting attribute features of the image of the vehicle to be identified can play a positive feedback role.
[0033] Secondly, embodiments of the present invention also provide a vehicle monitoring device, the vehicle monitoring device comprising:
[0034] The first acquisition module is used to acquire images of the vehicle to be identified;
[0035] The first determining module is used to determine a first suspicious vehicle image from the vehicle image to be identified based on the vehicle image to be identified and the vehicle robbery case attribute features, wherein the vehicle robbery case attribute features are extracted from historical vehicle robbery case information.
[0036] The second determining module is used to determine a second suspicious vehicle image from the first suspicious vehicle image based on the vehicle robbery case location information, wherein the vehicle robbery case location information is determined by the historical vehicle robbery case information;
[0037] The third determination module is used to perform vehicle behavior recognition processing on the second suspicious vehicle image through a preset behavior recognition model, and determine the third suspicious vehicle image in the second suspicious vehicle image.
[0038] The management module is used to compare the features of the third suspicious vehicle image with the vehicle image in the criminal record, determine whether the vehicle in the third suspicious vehicle image is the vehicle used to commit the vehicle robbery, and manage the vehicle used to commit the vehicle robbery.
[0039] Thirdly, embodiments of the present invention provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in the vehicle monitoring method provided in embodiments of the present invention.
[0040] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the vehicle monitoring method provided in the embodiments of the invention.
[0041] In this embodiment of the invention, an image of a vehicle to be identified is acquired; based on the image of the vehicle to be identified and the attribute features of a vehicle robbery case, a first suspicious vehicle image is identified from the image of the vehicle to be identified; a second suspicious vehicle image is identified from the first suspicious vehicle image based on the location information of the vehicle robbery case, which is determined by historical vehicle robbery case information; the second suspicious vehicle image is processed by a preset behavior recognition model to identify vehicle behavior, and a third suspicious vehicle image is identified from the second suspicious vehicle image; the third suspicious vehicle image is compared with the vehicle image in the criminal record to determine whether the vehicle in the third suspicious vehicle image is the vehicle used in the vehicle robbery, and the vehicle used in the vehicle robbery is managed. By using the above method to monitor vehicles, suspicious vehicles can be quickly located, response speed can be accelerated, and the efficiency of monitoring vehicles involved in vehicle robberies can be improved. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0043] Figure 1 This is a flowchart of a vehicle monitoring method provided in an embodiment of the present invention;
[0044] Figure 2 This is a flowchart of another vehicle monitoring method provided in an embodiment of the present invention;
[0045] Figure 3 This is a schematic diagram of the structure of a vehicle monitoring device provided in an embodiment of the present invention;
[0046] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0048] like Figure 1 As shown, Figure 1 This is a flowchart of a vehicle monitoring method provided in an embodiment of the present invention. The vehicle monitoring method includes the following steps:
[0049] 101. Obtain the image of the vehicle to be identified.
[0050] In this embodiment of the invention, the above-mentioned vehicle monitoring method can be applied to an urban security management platform. The urban security management platform has functions such as data processing, data transmission and reception, and data storage, and can be built based on a server or server cluster. The server or server cluster can be an electronic device with functions such as image processing, image acquisition, and image recognition.
[0051] The aforementioned vehicle images to be identified can be obtained by taking and uploading photos through electronic devices attached to the aforementioned urban security management platform. Generally, since there is more than one such electronic device, the number of vehicle images to be identified can be at least one. These images can be taken at certain time intervals using front-end cameras or cameras set up at intersections and traffic lights. The captured images can include images of motorcycles. The aforementioned time interval can be determined based on the frequency of vehicle robbery cases in the area where the electronic devices are deployed. Generally, the higher the frequency of vehicle robbery cases, the shorter the corresponding time interval should be.
[0052] The aforementioned vehicles can refer to any vehicle, whether motor or non-motorized, that can be used to commit vehicle hijacking, such as motorcycles, electric bicycles, and electric vehicles.
[0053] In this embodiment of the invention, the aforementioned urban security management platform uses an auxiliary electronic device to intermittently photograph vehicles in the area or on the street, obtains the photographed image data, and stores it in a database.
[0054] 102. Based on the images of the vehicles to be identified and the characteristics of vehicle robbery cases, the first suspicious vehicle image is identified from the images of the vehicles to be identified.
[0055] In this embodiment of the invention, the aforementioned first suspicious vehicle image may include, but is not limited to, motorcycle images and images of various motor vehicles and non-motor vehicles. The aforementioned vehicle robbery cases may include, but are not limited to, snatch-and-grab robberies, motor vehicle robberies, and motor vehicle violence incidents.
[0056] The aforementioned vehicle robbery case attribute features can be extracted from historical vehicle robbery case information to serve as the vehicle robbery case attribute features. Specifically, this historical vehicle robbery case information can be extracted from the case database of local public security organs or case-filing agencies. Generally, the case information in this database consists of historical cases that have already occurred. Case information can be queried based on specific case attributes, such as searching for vehicle robbery case information using the keyword "vehicle robbery." This historical vehicle robbery case information can include the location of the incident, the latitude and longitude of the corresponding location, and the corresponding distance, and includes, but is not limited to, the physical characteristics of the perpetrators and the physical characteristics of the tools used in the crime. The aforementioned vehicle robbery case attribute features can also refer to case information extracted from the attribute features of the aforementioned drive-by robbery cases, drive-by incidents, and drive-by violent incidents. For example, extracting attribute features from drive-by robbery cases can yield information such as two people wearing helmets, clothing attributes, carried items, and vehicle characteristics.
[0057] The aforementioned characteristics of vehicle robbery cases may include, but are not limited to, two people wearing two helmets, clothing attributes, carried items, and vehicle characteristics. Among these, two people wearing two helmets, clothing attributes, and vehicle characteristics are prerequisites for vehicle robbery. Generally speaking, if a vehicle image contains all three of these prerequisite attributes, it is suspected that the vehicle in the image may be involved in a vehicle robbery, and therefore it can be designated as the first suspicious vehicle.
[0058] In this embodiment of the invention, the urban security management platform performs vehicle attribute feature recognition on the collected images of vehicles to be identified, obtains the vehicle attribute features of the vehicles to be identified, and compares them with the attribute features of the vehicle robbery case. The vehicles to be identified that are successfully matched are set as the first suspicious vehicles.
[0059] 103. Based on the location information of the vehicle robbery case, the second suspicious vehicle image is determined from the first suspicious vehicle image.
[0060] In this embodiment of the invention, the second suspicious vehicle image can be obtained from the first suspicious vehicle image. The second suspicious vehicle can be a vehicle obtained by locating and confirming the vehicle in the first suspicious vehicle image using location information from a vehicle robbery case.
[0061] The location information of the aforementioned vehicle robbery cases can be obtained by extracting information from the aforementioned historical vehicle robbery cases. Specifically, by processing and analyzing information such as the location of the crime, the latitude and longitude of the corresponding location, and the corresponding distance in the historical vehicle robbery case information, the frequency of the crime in each region can be obtained. The first suspicious vehicle image obtained in the high crime frequency region is set as the second suspicious vehicle image, so that the vehicle in the first suspicious vehicle image can be spatially judged. For spatial judgment, a spatial area needs to be defined to determine whether the vehicle in the first suspicious vehicle image appears in the spatial area. The aforementioned spatial area can be divided based on the experience of the management personnel or based on the aforementioned historical vehicle robbery case information.
[0062] In one possible embodiment, the aforementioned urban security management platform extracts the location information of historical vehicle robbery cases and filters the obtained first suspicious vehicle image based on the location information to determine the second suspicious vehicle image.
[0063] In another possible embodiment, the aforementioned urban security management platform calculates the spatial area by weighting the areas divided by the managers based on experience and the locations of historical vehicle robbery cases, and sets the first suspicious vehicle image as the second suspicious vehicle for vehicles located in the first suspicious vehicle image within the aforementioned spatial area.
[0064] 104. By using a preset behavior recognition model to perform vehicle behavior recognition processing on the second suspicious vehicle image, a third suspicious vehicle image is identified from the second suspicious vehicle image.
[0065] In this embodiment of the invention, the third suspicious vehicle image can be obtained from the second suspicious vehicle image. The third suspicious vehicle can be a vehicle whose behavior is successfully identified by the preset behavior recognition model in the second suspicious vehicle image.
[0066] The aforementioned preset behavior recognition model can refer to a network model capable of recognizing vehicle loitering behavior. Specifically, the preset behavior recognition model can count the number of times a target vehicle appears based on a preset time interval and the number of electronic devices deployed in the deployment area, and analyze the behavior of the target vehicle based on the number of appearances to determine whether the target vehicle is loitering in the area. Generally speaking, if a vehicle with the aforementioned characteristics of vehicle robbery cases loiters in an area with a high frequency of such cases, there may be a possibility of vehicle robbery, and therefore it is necessary to monitor the target vehicle.
[0067] The aforementioned third suspicious vehicle image can refer to the image obtained after filtering the vehicles in the aforementioned second suspicious vehicle image through the aforementioned preset behavior recognition model to identify loitering behavior. Specifically, by performing data clustering on the vehicle data in the second suspicious vehicle image, the dwell data of the vehicles in the second suspicious vehicle image can be obtained. Finally, based on the dwell data, it is determined whether the vehicles in the second suspicious vehicle image conform to loitering behavior, and the second suspicious vehicle image that conforms is set as the third suspicious vehicle image. The aforementioned dwell data can refer to the basis for judging whether the vehicles in the second suspicious vehicle image are loitering in a certain area, and the determination of whether the vehicle has loitering behavior is based on the dwell data.
[0068] The aforementioned data clustering process can refer to the unified analysis and processing of data types with the same characteristics. For example, a unified loitering behavior analysis can be performed on images containing a second suspicious vehicle, and the second suspicious vehicle in the image that matches the loitering behavior can be identified as the third suspicious vehicle.
[0069] In one possible embodiment, the aforementioned urban security management platform identifies the loitering behavior of the second suspicious vehicle using a preset behavior recognition model, obtains the third suspicious vehicle, and stores the vehicle characteristics and driver characteristics of the third suspicious vehicle in the database.
[0070] 105. Compare the features of the third suspicious vehicle image with the vehicle image in the criminal record to determine whether the vehicle in the third suspicious vehicle image is the vehicle used to commit the vehicle robbery, and manage the vehicle used to commit the vehicle robbery.
[0071] In this embodiment of the invention, the aforementioned vehicle images with criminal records may refer to vehicle images recorded in the aforementioned historical vehicle robbery case information. Generally, the aforementioned urban public security management platform can obtain vehicle information of vehicles in the vehicle images with criminal records by reading case database information in the aforementioned local public security organs or case-filing agencies, and obtain vehicle characteristics of vehicles in the aforementioned vehicle images with criminal records by extracting the vehicle information of vehicles in the aforementioned vehicle images with criminal records.
[0072] The vehicle used to commit the vehicle robbery can refer to the third suspicious vehicle that was successfully compared with the vehicle characteristics in the above-mentioned vehicle image in the criminal case file and the vehicle characteristics in the above-mentioned third suspicious vehicle image. Specifically, it can be any vehicle capable of committing the vehicle robbery, such as a motor vehicle or a non-motor vehicle, such as a motorcycle or an electric bicycle.
[0073] The aforementioned management may refer to the process of storing the vehicle features of the third suspicious vehicle image obtained through successful comparison into the case database of the local public security organ or case-filing agency, or it may refer to the process of optimizing the vehicle features of the third suspicious vehicle image that failed to be compared, as well as the method parameters in the acquisition process.
[0074] In one possible embodiment, after a successful match between the vehicle involved in the robbery and the vehicle in the criminal record vehicle image, and it is found that the criminal record vehicle has been captured based on the case information, the vehicle features of the vehicle involved in the robbery are structurally broken down to obtain more detailed vehicle features, such as tire wear, and then compared again with the vehicle in the criminal record vehicle image. If the comparison is still successful, the information is sent to the display terminal corresponding to the urban security management platform for display to the management personnel for manual review. The display terminal can be any mobile device capable of data processing and transmission, such as a mobile phone or laptop.
[0075] In another possible embodiment, if the vehicle used to commit the vehicle robbery fails to match the vehicle in the criminal record, the vehicle features of the vehicle used to commit the vehicle robbery are structurally decomposed to obtain more detailed vehicle features, which are then compared with the vehicle features of the vehicle in the criminal record vehicle image.
[0076] In this embodiment of the invention, an image of a vehicle to be identified is acquired; based on the image of the vehicle to be identified and the attribute features of a vehicle robbery case, a first suspicious vehicle image is identified from the image of the vehicle to be identified; a second suspicious vehicle image is identified from the first suspicious vehicle image based on the location information of the vehicle robbery case, which is determined by historical vehicle robbery case information; the second suspicious vehicle image is processed by a preset behavior recognition model to identify vehicle behavior, and a third suspicious vehicle image is identified from the second suspicious vehicle image; the third suspicious vehicle image is compared with the vehicle image in the criminal record to determine whether the vehicle in the third suspicious vehicle image is the vehicle used in the vehicle robbery, and the vehicle used in the vehicle robbery is managed. By using the above method to monitor vehicles, suspicious vehicles can be quickly located, response speed can be accelerated, and the efficiency of monitoring vehicles involved in vehicle robberies can be improved.
[0077] Optionally, in the step of determining the first suspicious vehicle image from the images of vehicles to be identified based on the images of vehicles to be identified and the attribute features of vehicle robbery cases, each image of vehicles to be identified can be identified to obtain the images of each target vehicle in each image of vehicles to be identified. Then, the image of each target vehicle is used to extract features from the image of each target vehicle to obtain the attribute features of each target vehicle. Finally, the attribute features of each target vehicle are compared with the attribute features of vehicle robbery cases. If the attribute features of the target vehicle are successfully compared with the attribute features of vehicle robbery cases, the image of the target vehicle that is successfully compared is determined as the first suspicious vehicle image.
[0078] In this embodiment of the invention, the aforementioned structured engine can refer to an image processing technique that extracts features from the obtained image using a network model. Specifically, the structured engine extracts vehicle features from each target vehicle in the image of the vehicle to be identified, obtaining the attribute features of each target vehicle, such as the driver's clothing, whether a helmet is worn, vehicle color, vehicle structure, and other information. The aforementioned target vehicle can refer to a vehicle in the image of the vehicle to be identified that can be identified by the structured engine and whose corresponding attribute features can be obtained. An image of one target vehicle may include at least one target vehicle.
[0079] Specifically, the attribute features of the target vehicle obtained above can be compared with the attribute features of vehicle robbery cases in the case database of the local public security bureau or case-filing agency. The similarity is compared to a similarity threshold. When the similarity is greater than or equal to the threshold, the target vehicle is considered the first suspect vehicle. The similarity threshold can be adjusted based on the number of vehicles (i.e., the target vehicle) that can be identified by the structured engine in the image of the vehicle to be identified. Generally, the more target vehicles there are, the greater the probability of similarity; therefore, the similarity threshold needs to be increased to ensure that the comparison result does not deviate from the actual situation. Generally, the similarity threshold will not exceed 1. The similarity can be Euclidean distance similarity or cosine similarity.
[0080] In one possible embodiment, the aforementioned urban security management platform performs target vehicle identification on the images of the vehicles to be identified, obtains images of each target vehicle, extracts features from the images of each target vehicle using a structured engine, obtains the attribute features of each target vehicle, and compares the similarity of the attribute features of each target vehicle with the attribute features of a vehicle robbery case, and identifies the target vehicle that successfully matches as the first suspicious vehicle.
[0081] Optionally, in the step before determining the image of the second suspicious vehicle based on the location information of the vehicle robbery case and the image of the first suspicious vehicle, the location information of the vehicle robbery case can be used to determine the locations near the case location, and then the latitude and longitude of the locations near the locations can be used to determine the area where the locations are concentrated, and the area where the locations are concentrated can be used as the target area.
[0082] In this embodiment of the invention, the location information of the aforementioned vehicle robbery case can be extracted from the case database of the local public security organ or case-filing agency through a network connection of the aforementioned urban public security management platform. This information includes, but is not limited to, the crime scene of the vehicle robbery, the addresses of relevant personnel, and the addresses where criminal evidence was found. The aforementioned location surrounding the case can refer to the specific coordinates of the crime scene of the vehicle robbery, which can be precisely measured in latitude and longitude. Generally, the latitude and longitude coordinates of the crime scene of the vehicle robbery can be confirmed using a map or GPS to obtain the latitude and longitude of the location surrounding the case. For example, by extracting the location information of the aforementioned snatch-and-grab robbery case, information such as the crime scene of the snatch-and-grab robbery, the addresses of relevant personnel, and the addresses where criminal evidence was found can be obtained.
[0083] The aforementioned target area can be determined by analyzing the latitude and longitude distribution of points near the locations of the aforementioned cases to identify the distribution of vehicle snatching cases. Then, based on this distribution, if the number of vehicle snatching cases within a certain area exceeds a warning threshold, that area is designated as the target area. Specifically, the scope of the target area can be dynamically adjusted based on the number of vehicle snatching cases and their distribution. Generally, the scope can be determined based on the minimum number of cases and the most concentrated case distribution. For example, if there are three cases, and the locations of the three cases are no more than 500 meters apart, the locations of these three cases are combined, and the overlapping area is designated as the target area. Similarly, the distribution of drive-by snatching cases can be analyzed to determine the target area.
[0084] Optionally, in the step of determining the second suspicious vehicle image based on the location information of the vehicle robbery case and the first suspicious vehicle image, if the first suspicious vehicle image has a license plate, a first count detection is performed in the target area using the license plate of the vehicle in the first suspicious vehicle image to obtain the number of times the vehicle in the first suspicious vehicle image appears in the target area. If the first suspicious vehicle image does not have a license plate, a second count detection is performed in the target area using the vehicle features of the vehicle in the first suspicious vehicle image to obtain the number of times the vehicle in the first suspicious vehicle image appears in the target area. Finally, based on the number of times the vehicle in the first suspicious vehicle image appears in the target area, the second suspicious vehicle image is determined from the first suspicious vehicle image.
[0085] In this embodiment of the invention, the aforementioned frequency detection can refer to statistically analyzing the number of times a vehicle in the first suspicious vehicle image appears within the target area using the data calculation model in the urban security management platform. Specifically, the first frequency detection can be based on the premise that the license plate of the vehicle in the first suspicious vehicle image can be identified. Then, the license plate information of the vehicle in the first suspicious vehicle image is statistically processed along with the number of times the same license plate appears in different times and locations within the target area to obtain the number of times the vehicle in the first suspicious vehicle image appears within the target area. The second frequency detection can be based on the premise that the license plate of the vehicle in the first suspicious vehicle image cannot be identified. Then, the vehicle features of the vehicle in the first suspicious vehicle image are statistically processed along with the number of times the same vehicle features appear in different times and locations within the target area to obtain the number of times the first suspicious vehicle appears within the target area.
[0086] In one possible embodiment, when performing the aforementioned frequency detection, a shooting time T needs to be set within the target area. The shooting time T can be adjusted based on the number of electronic devices in the target area, such as the number of electronic cameras, and is denoted as N. The aforementioned urban security management platform uses N electronic devices within the target area to continuously monitor the license plate or vehicle features of the vehicles in the first suspicious vehicle image for a period of T. The frequency statistics calculated from the license plate or vehicle features are entered into the platform database, and the frequently appearing first suspicious vehicle is designated as the second suspicious vehicle. Generally speaking, if a vehicle appears frequently in a certain area and exhibits the aforementioned characteristics of vehicle robbery cases, it is determined that the vehicle is likely to be involved in vehicle robbery and is therefore marked as the second suspicious vehicle.
[0087] Optionally, in the step of performing vehicle behavior recognition processing on the second suspicious vehicle image through a preset behavior recognition model to determine the third suspicious vehicle image, a preset clustering time can also be obtained. Then, based on the preset clustering time, a sliding window algorithm is used to count the vehicles in the second suspicious vehicle image within the target area to obtain the dwell data of the vehicles in the second suspicious vehicle image. The dwell data of the vehicles in the second suspicious vehicle image includes the number of times the vehicles in the second suspicious vehicle image are photographed in the target area and the vehicle images photographed during the dwell time. Finally, based on the preset dwell threshold and the number of dwell times, the third suspicious vehicle image is determined from the second suspicious vehicle image.
[0088] In this invention, the clustering time can be adjusted according to the size of the target area. Generally, the larger the target area, the longer the clustering time can be set to ensure sufficient data for clustering calculations and thus guarantee accuracy. The sliding window algorithm can use the clustering time as the window size, statistically analyze the captured image data for different time periods based on different clustering times, and obtain the vehicle's dwell time data in the second suspicious vehicle image. The dwell time data can refer to the number of times the vehicle in the second suspicious vehicle image is captured within the target area and the vehicle images captured during those dwell times. The preset dwell time threshold can be dynamically adjusted according to the size of the target area; generally, the larger the target area, the larger the preset dwell time threshold.
[0089] In one possible embodiment, the urban security management platform sets a corresponding clustering time based on the size of the target area. Generally, the clustering time is increased by 10 seconds for every 100m increase in the target area. The specific setting can be adjusted according to the specific implementation plan. Then, based on the clustering time, the number of times the second suspicious vehicle appears in the target area is counted. For example, the images of the second suspicious vehicle captured by all electronic devices are counted every ten seconds to obtain the number of times the vehicle stops in the second suspicious vehicle image. Then, the number of stops is compared with the preset stop threshold. If it is greater than or equal to the preset stop threshold, it is determined that the vehicle in the second suspicious vehicle image has a loitering behavior in the target area. Therefore, the second suspicious vehicle image is used as the third suspicious vehicle image.
[0090] Optionally, in the step of comparing the features of the third suspicious vehicle image with the vehicle image in the criminal record to determine whether the vehicle in the third suspicious vehicle image is the vehicle that committed the vehicle robbery, the license plate of the vehicle in the third suspicious vehicle image can be compared with the license plate of the vehicle in the criminal record to obtain a first comparison result. Then, the vehicle features of the vehicle in the third suspicious vehicle image can be compared with the vehicle features of the vehicle in the criminal record to obtain a second comparison result. Finally, based on at least one of the first comparison result and the second comparison result, it can be determined whether the third suspicious vehicle is the vehicle that committed the vehicle robbery.
[0091] In this embodiment of the invention, by comparing the license plate or vehicle features of a vehicle with the license plate or vehicle features of a third suspicious vehicle image, the vehicle in the third suspicious vehicle image that is successfully matched is identified as the vehicle that committed the vehicle robbery and is entered into the case database.
[0092] Specifically, if the vehicles in the aforementioned third suspicious vehicle image were counted based on their license plates during the clustering data calculation, then in this embodiment, the license plates are also compared with those of vehicles with prior criminal records to obtain a first comparison result. If the first comparison result is successful, then the third suspicious vehicle is treated as the vehicle involved in the vehicle robbery.
[0093] If the vehicles in the aforementioned third suspicious vehicle image were counted based on vehicle features during the clustering data calculation, then in this embodiment, the vehicle features are also compared with the vehicle features of vehicles with criminal records to obtain a second comparison result. If the second comparison result is a successful comparison, then the vehicles in the third suspicious vehicle image that were successfully compared are treated as vehicles that committed vehicle robbery.
[0094] In one possible embodiment, if the vehicle in the aforementioned third suspicious vehicle image has both license plate and vehicle features, a weighted algorithm will be used to calculate the first comparison result obtained through license plate comparison and the second comparison result obtained through vehicle feature comparison to obtain the total comparison result. Based on the total comparison result, it will be determined whether the vehicle in the aforementioned third suspicious vehicle image is the vehicle that committed the vehicle robbery. Specifically, due to the uniqueness of license plates, license plates generally have a weight of 0.7 in the weighted calculation, while vehicle features have a weight of 0.3. For example, when the aforementioned third suspicious vehicle is compared with the license plates of 10 vehicles with criminal records through license plate comparison, and 7 license plate images are successfully compared, while when the vehicle features are compared with the images of 10 vehicles with criminal records, and all 10 images are unsuccessful, then according to the weight ratio calculation, it will still be determined that the vehicle in the third suspicious vehicle image is successfully compared with the vehicles with criminal records.
[0095] More specifically, the overall comparison result is compared with the comparison threshold. If the comparison result is greater than or equal to the comparison threshold, the vehicle in the third suspicious vehicle image is determined to have the same feature attribute as the vehicle in the criminal record, and is processed as the vehicle involved in the vehicle robbery. The specific comparison threshold can be set according to the specific clustering calculation time.
[0096] Optionally, in the step of comparing the features of the vehicle in the third suspicious vehicle image with the vehicle in the case file to determine whether the vehicle in the third suspicious vehicle image is the vehicle that committed the vehicle robbery, and managing the vehicle that committed the vehicle robbery, if it is determined that the vehicle in the third suspicious vehicle image is the vehicle that committed the vehicle robbery, then the vehicle features of the vehicle in the third suspicious vehicle image are input into the database and the case is recorded, so that the vehicle that committed the vehicle robbery can be tracked later. If it is determined that the vehicle in the third suspicious vehicle image is not the vehicle that committed the vehicle robbery, then the vehicle features of the vehicle that committed the vehicle robbery are treated as failure attributes, so that the method for extracting the attribute features of the vehicle image to be identified has a positive feedback effect.
[0097] In this embodiment of the invention, the aforementioned follow-up tracking can refer to extracting the case attributes corresponding to the criminal vehicle, and then using electronic devices attached to the aforementioned urban security management platform to investigate the criminal vehicle based on the case attributes. The aforementioned failure attributes can be parameters used by the aforementioned structured engine when extracting attribute features, that is, the pre-parameters when the comparison between the vehicle in the third suspicious vehicle image and the vehicle with the criminal record fails. Based on these pre-parameters, the feature extraction method of the aforementioned structured engine is optimized, thereby making the feature extraction of suspicious vehicles by the structured engine more accurate.
[0098] In one possible embodiment, regardless of whether the vehicle in the third suspicious vehicle image is successfully matched with the vehicle in the criminal record, it is necessary to conduct a manual review to compare and determine the additional vehicle features of the vehicle in the third suspicious vehicle image, so as to determine whether the vehicle in the third suspicious vehicle image really belongs to the vehicle used to commit the vehicle robbery. The additional vehicle features can be vehicle features that can highlight the characteristics of vehicle crime, such as handbags, clubs, etc., which cannot be identified by the structured engine.
[0099] In another possible embodiment, when the comparison between the vehicle in the third suspicious vehicle image and the vehicle in the criminal record fails, the clustering time can be shortened during the optimization process, and then the clustering calculation can be performed again. For example, the initial 30 minutes can be reduced to 15 minutes, 10 minutes or 5 minutes to perform the clustering calculation once, so that the obtained data is more concentrated and no useful data features are missed. This optimizes the process of vehicle feature extraction by the structured engine.
[0100] In another possible embodiment, when the vehicle in the third suspicious vehicle image fails to match the vehicle in the criminal record, the scope of the target area can be reduced during the optimization process, from the entire target area to 1 / 4 of the target area, thereby reducing the traffic flow of the vehicle identification, eliminating interference points, and increasing the identification accuracy.
[0101] like Figure 2 As shown in the figure, an embodiment of the present invention also provides a flowchart of another vehicle monitoring method, characterized in that it includes:
[0102] Images captured by the front-end camera are sent to a structured image engine. Based on the characteristics of vehicle robbery crimes, motorcycle attributes are extracted from the captured images to obtain the attribute features of each target motorcycle, including: two people wearing helmets, clothing color, carried items, motorcycle color, license plate, and other motorcycle characteristics. It should be noted that two people wearing helmets, clothing color, and motorcycle color are prerequisites; only if the two people wearing helmets, motorcycle color, and clothing color are dark can the next step of judgment be performed. Subsequently, a key area filtering method is used to filter the target motorcycles. These key areas can be manually defined or derived from vehicle robbery reports received by the police, extracting attributes from the cases to obtain the locations near the crime scenes. Then, the distances between these locations are calculated based on their latitude and longitude, grouping similar locations together as a key area. Finally, the target motorcycles are filtered based on these area locations. After filtering through key areas, the filtered data is divided into those with license plate features and those without, and compared and calculated accordingly. (That is, the number of times a motorcycle appears on N devices within a time period T is counted to obtain the number of appearances Q, and Q is compared with the appearance threshold to determine whether a suspected motorcycle is a suspected motorcycle). Then, the suspected motorcycles are subjected to data clustering calculation (i.e., the clustering calculation method and setting process are the same as the above clustering calculation). Suspected motorcycles identified as loitering motorcycles are compared with license plates or features in the feature database. Regardless of whether the comparison is successful or unsuccessful, a second manual review is required. By comparing the changes in the characteristics of the motorcycle's carried over time, it is determined whether the loitering motorcycle is the one that committed the vehicle robbery. If the comparison of loitering motorcycles fails, the feature extraction model for loitering motorcycles is optimized. The specific optimization method can be obtained from the following:
[0103] 1. Shorten the time data, such as to 30 minutes, recalculate the occurrence count Q, and then adjust the capture time clustering time, such as setting it to 15, 20, or 25 seconds, to obtain the number of hovering times. Adjust each image multiple times to obtain a relatively accurate time T. Perform the same operation on each image to obtain all times T1, T2, T3... Finally, obtain a median and verify it. If the accuracy is good, apply it to the actual algorithm; if the accuracy is poor, continue adjusting until a relatively accurate value is obtained.
[0104] 2. Selection of statistical areas: By querying video capture data of all devices in the area, the area is divided into areas with high and low traffic. The proportions of points with high traffic, low traffic, and high and low traffic points are calculated separately. The total number of occurrences and time clustering data under various conditions are obtained, thereby dividing the median value. This allows for more reasonable division of areas and elimination of interfering points.
[0105] If the aforementioned loitering vehicle is successfully matched with a vehicle in the criminal record, it will be added to the database and imported into a new case.
[0106] like Figure 3 As shown, this embodiment of the invention also provides a vehicle monitoring device, characterized in that it includes:
[0107] The first acquisition module 301 is used to acquire an image of the vehicle to be identified;
[0108] The first determining module 302 is used to determine a first suspicious vehicle image in the vehicle image to be identified based on the vehicle image to be identified and the vehicle robbery case attribute features, wherein the vehicle robbery case attribute features are extracted from historical vehicle robbery case information.
[0109] The second determining module 303 is used to determine a second suspicious vehicle image from the first suspicious vehicle image based on the vehicle robbery case location information, wherein the vehicle robbery case location information is determined by the historical vehicle robbery case information;
[0110] The third determining module 304 is used to perform vehicle behavior recognition processing on the second suspicious vehicle image through a preset behavior recognition model, and determine the third suspicious vehicle image in the second suspicious vehicle image.
[0111] The management module 305 is used to compare the features of the third suspicious vehicle image with the vehicle image in the criminal record, determine whether the vehicle in the third suspicious vehicle image is the vehicle that committed the vehicle robbery, and manage the vehicle that committed the vehicle robbery.
[0112] Optionally, the first determining module 302 mentioned above includes:
[0113] The first acquisition submodule is used to identify each of the images of vehicles to be identified, and to obtain an image of each target vehicle in each of the images of vehicles to be identified, wherein an image of a target vehicle includes at least one target vehicle;
[0114] The first extraction submodule is used to extract features from the images of each target vehicle using a structured engine to obtain the attribute features of each target vehicle.
[0115] The first comparison submodule is used to compare the attribute features of each target vehicle with the attribute features of the vehicle robbery case;
[0116] The second comparison submodule is used to identify the image of the target vehicle as the first suspicious vehicle image if the attribute features of the target vehicle are successfully compared with the attribute features of the vehicle robbery case.
[0117] Optionally, the above-mentioned device further includes:
[0118] The fourth determining module is used to determine the locations near the crime scene based on the location information of the vehicle robbery case.
[0119] The fifth determining module is used to determine the area where the points are concentrated based on the latitude and longitude of the attached points, and to take the area where the points are concentrated as the target area.
[0120] Optionally, the second determining module 303 further includes:
[0121] The second acquisition submodule is used to perform a first count detection on the target area using the license plate of the vehicle in the first suspicious vehicle image if the first suspicious vehicle image has a license plate, and obtain the number of times the vehicle in the first suspicious vehicle image appears in the target area.
[0122] The third acquisition submodule is used to perform a second number detection in the target area based on the vehicle features of the vehicle in the first suspicious vehicle image if the first suspicious vehicle image does not have a license plate, so as to obtain the number of times the vehicle in the first suspicious vehicle image appears in the target area.
[0123] The first determining submodule is used to determine a second suspicious vehicle image from the first suspicious vehicle image based on the number of times the vehicle appears in the target area in the first suspicious vehicle image.
[0124] Optionally, the third determining module 304 mentioned above further includes:
[0125] The fourth acquisition submodule is used to acquire a preset clustering time, which can be adjusted according to the size of the target region;
[0126] The fifth acquisition submodule is used to statistically analyze the vehicles in the second suspicious vehicle images in the target area based on the preset clustering time using a sliding window algorithm, and obtain the dwell data of the vehicles in the second suspicious vehicle images. The dwell data includes the number of times the vehicles in the second suspicious vehicle images are captured in the target area and the vehicle images captured during the dwell time.
[0127] The second determination submodule is used to determine a third suspicious vehicle image from the second suspicious vehicle image based on a preset dwell threshold and the number of dwell times.
[0128] Optionally, the management module 305 mentioned above includes:
[0129] The second comparison submodule is used to compare the license plate of the vehicle in the third suspicious vehicle image with the license plate of the vehicle with the criminal record to obtain the first comparison result;
[0130] The third comparison submodule is used to compare the vehicle features of the vehicle in the third suspicious vehicle image with the vehicle features of the vehicle with the criminal record to obtain a second comparison result;
[0131] The judgment submodule is used to determine, based on at least one of the first comparison result and the second comparison result, whether the vehicle in the third suspicious vehicle image is the vehicle that committed the vehicle robbery.
[0132] Optionally, the management module 305 mentioned above also includes:
[0133] The tracking submodule is used to input the vehicle features in the third suspicious vehicle image into the database and record the incident case if it is determined that the vehicle in the third suspicious vehicle image is the vehicle that committed the vehicle robbery, so that the vehicle that committed the vehicle robbery can be tracked later.
[0134] The feedback submodule is used to, if it is determined that the vehicle in the third suspicious vehicle image is not the vehicle that committed the vehicle robbery, then treat the vehicle features of the vehicle that committed the vehicle robbery as a failure attribute, so that the method for extracting the attribute features of the vehicle image to be identified can play a positive feedback role.
[0135] like Figure 4 As shown, an embodiment of the present invention also provides an electronic device, characterized in that it includes a processor, which can execute any of the above-described vehicle monitoring methods.
[0136] Specifically, it includes a processor 401 and a memory 402, as well as a computer program stored in the memory 402 and capable of running on the processor 401 to execute the vehicle monitoring method, wherein:
[0137] The processor 401 executes the calculator program for the vehicle monitoring method stored in the memory 402, performing the following steps:
[0138] Acquire an image of the vehicle to be identified;
[0139] Based on the vehicle image to be identified and the vehicle robbery case attribute features, a first suspicious vehicle image is determined from the vehicle image to be identified, and the vehicle robbery case attribute features are extracted from historical vehicle robbery case information;
[0140] Based on the location information of the vehicle robbery case, a second suspicious vehicle image is determined from the first suspicious vehicle image, wherein the location information of the vehicle robbery case is determined by the historical vehicle robbery case information;
[0141] The second suspicious vehicle image is processed by a preset behavior recognition model to identify the vehicle behavior, and a third suspicious vehicle image is determined from the second suspicious vehicle image.
[0142] The third suspicious vehicle image is compared with the vehicle image in the criminal record to determine whether the vehicle in the third suspicious vehicle image is the vehicle used to commit the vehicle robbery, and the vehicle used to commit the vehicle robbery is then managed.
[0143] Optionally, the vehicle monitoring method is characterized in that the processor 401 performs the following: based on the image of the vehicle to be identified and the attribute features of the vehicle robbery case, it determines a first suspicious vehicle image from the image of the vehicle to be identified, including:
[0144] Each of the images of vehicles to be identified is identified to obtain an image of each target vehicle in each of the images of vehicles to be identified, wherein an image of a target vehicle includes at least one target vehicle.
[0145] The structured engine is used to extract features from the images of each target vehicle to obtain the attribute features of each target vehicle.
[0146] The attribute features of each target vehicle are compared with the attribute features of the vehicle robbery case. If the attribute features of the target vehicle are successfully matched with the attribute features of the vehicle robbery case, the image of the target vehicle that is successfully matched is identified as the first suspicious vehicle image.
[0147] Optionally, the vehicle monitoring method is characterized in that, before the processor 401 executes the step of determining the second suspicious vehicle image based on the location information of the vehicle robbery case and the first suspicious vehicle image, it includes:
[0148] Based on the location information of the vehicle robbery case, determine the locations near the case location;
[0149] Based on the latitude and longitude of the attached points, determine the area where the points are concentrated, and use the area where the points are concentrated as the target area.
[0150] Optionally, the vehicle monitoring method is characterized in that the processor 401 executes the step of determining a second suspicious vehicle image based on the location information of the vehicle robbery case and the first suspicious vehicle image, including:
[0151] If the first suspicious vehicle image contains a license plate, then the first count detection is performed in the target area using the license plate of the vehicle in the first suspicious vehicle image to obtain the number of times the vehicle in the first suspicious vehicle image appears in the target area;
[0152] If the first suspicious vehicle image does not have a license plate, a second number detection is performed in the target area based on the vehicle features of the vehicle in the first suspicious vehicle image to obtain the number of times the vehicle in the first suspicious vehicle image appears in the target area;
[0153] Based on the number of times the first suspicious vehicle appears in the target area, a second suspicious vehicle image is determined from the first suspicious vehicle image.
[0154] Optionally, the vehicle monitoring method is characterized in that the processor 401 performs vehicle behavior recognition processing on the second suspicious vehicle image using a preset behavior recognition model to determine the third suspicious vehicle image, including:
[0155] A preset clustering time is obtained, which can be adjusted according to the size of the target region;
[0156] Based on the preset clustering time, the sliding window algorithm is used to count the vehicles in the second suspicious vehicle images in the target area to obtain the dwell data of the second suspicious vehicle. The dwell data includes the number of times the vehicle in the second suspicious vehicle image is captured in the target area and the vehicle image captured during the dwell time.
[0157] Based on a preset dwell threshold and the number of dwell times, a third suspicious vehicle image is determined from the second suspicious vehicle image.
[0158] Optionally, the vehicle monitoring method is characterized in that the processor 401 performs feature comparison between the third suspicious vehicle image and the vehicle image in the criminal record to determine whether the vehicle in the third suspicious vehicle image is the vehicle used to commit vehicle robbery, including:
[0159] The license plate of the vehicle in the third suspicious vehicle image is compared with the license plate of the vehicle in the case record vehicle image to obtain the first comparison result;
[0160] The vehicle features in the third suspicious vehicle image are compared with the vehicle features in the vehicle in the criminal record image to obtain a second comparison result;
[0161] Based on at least one of the first comparison result and the second comparison result, determine whether the third suspicious vehicle is the vehicle used to commit vehicle robbery.
[0162] Optionally, the vehicle monitoring method is characterized in that, in the process of processor 401 managing the vehicle involved in the vehicle hijacking, it further includes:
[0163] If it is determined that the vehicle in the third suspicious vehicle image is the vehicle that committed the vehicle robbery, then the vehicle characteristics of the vehicle in the third suspicious vehicle image are entered into the database and the case is recorded so that the vehicle that committed the vehicle robbery can be tracked later.
[0164] If it is determined that the third suspicious vehicle is not the vehicle that committed the vehicle robbery, then the vehicle features of the vehicle that committed the vehicle robbery are taken as failure attributes, so that the method for extracting attribute features of the image of the vehicle to be identified can play a positive feedback role.
[0165] This invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the vehicle monitoring method or the application-side vehicle monitoring method provided in this invention, and achieves the same technical effect. To avoid repetition, it will not be described again here.
[0166] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0167] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.
Claims
1. A vehicle monitoring method, characterized in that, The method includes the following steps: Acquire an image of the vehicle to be identified; Based on the vehicle image to be identified and the vehicle robbery case attribute features, a first suspicious vehicle image is determined from the vehicle image to be identified, and the vehicle robbery case attribute features are extracted from historical vehicle robbery case information; Based on the location information of the vehicle robbery case, a second suspicious vehicle image is determined from the first suspicious vehicle image, wherein the location information of the vehicle robbery case is determined by the historical vehicle robbery case information; The second suspicious vehicle image is processed by a preset behavior recognition model to identify the vehicle behavior, and a third suspicious vehicle image is determined from the second suspicious vehicle image. The third suspicious vehicle image is compared with the vehicle image in the criminal record to determine whether the vehicle in the third suspicious vehicle image is the vehicle used to commit the vehicle robbery, and the vehicle used to commit the vehicle robbery is then managed.
2. The vehicle monitoring method as described in claim 1, characterized in that, The step of identifying a first suspicious vehicle image from the vehicle images to be identified, based on the images of the vehicle to be identified and the attribute features of vehicle robbery cases, includes: The image of the vehicle to be identified is identified to obtain images of each target vehicle in the image of the vehicle to be identified, and the image of each target vehicle includes at least one target vehicle. The structured engine is used to extract features from the images of each target vehicle to obtain the attribute features of each target vehicle. The attribute characteristics of each target vehicle are compared with the attribute characteristics of the vehicle robbery case; If the attribute characteristics of the target vehicle are successfully matched with the attribute characteristics of the vehicle robbery case, then the image of the target vehicle that is successfully matched is identified as the first suspicious vehicle image.
3. The vehicle monitoring method as described in claim 2, characterized in that, Before determining the image of the second suspicious vehicle based on the location information of the vehicle robbery case and the image of the first suspicious vehicle, the process includes: Based on the location information of the vehicle robbery case, determine the locations near the case location; Based on the latitude and longitude of the aforementioned attachment points, determine the area where the points are concentrated; The area where the points are concentrated is taken as the target area.
4. The vehicle monitoring method as described in claim 3, characterized in that, The step of determining the image of the second suspicious vehicle based on the location information of the vehicle robbery case and the image of the first suspicious vehicle includes: If the first suspicious vehicle image contains a license plate, then the first count detection is performed in the target area using the license plate of the vehicle in the first suspicious vehicle image to obtain the number of times the vehicle in the first suspicious vehicle image appears in the target area; If the first suspicious vehicle image does not have a license plate, a second number detection is performed in the target area based on the vehicle features of the vehicle in the first suspicious vehicle image to obtain the number of times the vehicle in the first suspicious vehicle image appears in the target area; Based on the number of times the vehicle appears in the target area in the first suspicious vehicle image, a second suspicious vehicle image is determined from the first suspicious vehicle image.
5. The vehicle monitoring method as described in any one of claims 2 to 4, wherein the step of performing vehicle behavior recognition processing on the second suspicious vehicle image using a preset behavior recognition model to determine the third suspicious vehicle image includes: A preset clustering time is obtained, which can be adjusted according to the size of the target region; Based on the preset clustering time, a sliding window algorithm is used to count the vehicles in the second suspicious vehicle images within the target area to obtain the dwell data of the vehicles in the second suspicious vehicle images. The dwell data includes the number of times the vehicles in the second suspicious vehicle images were photographed in the target area and the vehicle images photographed during the dwell time. Based on a preset dwell threshold and the number of dwell times, a third suspicious vehicle image is determined from the second suspicious vehicle image.
6. The vehicle monitoring method as described in claim 5, characterized in that, The step of comparing the features of the third suspicious vehicle image with the vehicle image in the criminal record to determine whether the vehicle in the third suspicious vehicle image is the vehicle used in the vehicle robbery includes: The license plate of the vehicle in the third suspicious vehicle image is compared with the license plate of the vehicle in the case record vehicle image to obtain the first comparison result; The vehicle features in the third suspicious vehicle image are compared with the vehicle features in the vehicle image with the criminal record, to obtain a second comparison result; Based on at least one of the first comparison result and the second comparison result, determine whether the vehicle in the third suspicious vehicle image is the vehicle used to commit vehicle robbery.
7. The vehicle monitoring method as described in claim 6, characterized in that, The management of the vehicles involved in the vehicle hijacking also includes: If it is determined that the vehicle in the third suspicious vehicle image is the vehicle used to commit the vehicle robbery, then the vehicle characteristics in the third suspicious vehicle image are entered into the database and the case is recorded so that the vehicle used to commit the vehicle robbery can be tracked subsequently. If it is determined that the vehicle in the third suspicious vehicle image is not the vehicle that committed the vehicle robbery, then the vehicle features of the vehicle that committed the vehicle robbery are taken as failure attributes, so that the method for extracting the attribute features of the vehicle image to be identified can play a positive feedback role.
8. A vehicle monitoring device, characterized in that, The traction component monitoring device includes: The first acquisition module is used to acquire images of the vehicle to be identified; The first determining module is used to determine a first suspicious vehicle image from the vehicle image to be identified based on the vehicle image to be identified and the vehicle robbery case attribute features, wherein the vehicle robbery case attribute features are extracted from historical vehicle robbery case information. The second determining module is used to determine a second suspicious vehicle image from the first suspicious vehicle image based on the vehicle robbery case location information, wherein the vehicle robbery case location information is determined by the historical vehicle robbery case information; The third determination module is used to perform vehicle behavior recognition processing on the second suspicious vehicle image through a preset behavior recognition model, and determine the third suspicious vehicle image in the second suspicious vehicle image. The management module is used to compare the features of the third suspicious vehicle image with the vehicle image in the criminal record, determine whether the vehicle in the third suspicious vehicle image is the vehicle used to commit the vehicle robbery, and manage the vehicle used to commit the vehicle robbery.
9. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the vehicle monitoring method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the vehicle monitoring method as described in any one of claims 1 to 7.