Traffic violation education management method, computer equipment and readable storage medium

Abstract

The invention discloses a traffic violation education management method, computer equipment and a readable storage medium, and the method comprises the steps: inputting a monitoring image or video into a traffic violation detection model, so as to recognize traffic violation behavior data or a data set; performing face and license plate recognition processing on the violation image or the violation frame image to obtain a face recognition result and a license plate recognition result; whether the face recognition result is invalid face data or not and whether the license plate recognition result is invalid license plate data or not are judged, and if not, identity recognition calibration processing is carried out according to the face recognition result and / or the license plate recognition result, and illegal people are educated; and if yes, carrying out identity checking processing according to the license plate RFID information group and carrying out education processing on violation personnel. According to the invention, information of violation personnel can be determined by adopting a plurality of identity recognition modes and educated and processed, and the detection efficiency and accuracy are high, so that the manual patrol workload of traffic police is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic violation technology, and in particular to a traffic violation education and management method, a computer device, and a readable storage medium. Background Art

[0002] With the increasing popularity and number of motorcycles and electric bicycles, violations of road traffic regulations involving them are also increasing, posing a potential safety hazard. To effectively reduce electric vehicle traffic violations and improve both personal and road safety, traffic management departments in various cities currently deploy police officers at intersections to inspect passing motorcycles and electric bicycles and provide on-site education to offending drivers. However, this approach to traffic education and management is plagued by issues such as insufficient traffic police personnel, insufficient budgets, and a lack of a long-term mechanism for traffic management. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a traffic violation education management method, computer equipment and readable storage medium, which can use multiple identity recognition methods to determine the information of violators and educate them, with high detection efficiency and accuracy, thereby reducing the workload of traffic police's manual patrols.

[0004] In order to solve the above technical problems, the present invention provides a traffic violation education and management method, comprising: obtaining surveillance images or videos of a traffic area, and a license plate RFID collection database; inputting the surveillance images or videos into a traffic violation detection model to identify traffic violation behavior data or data groups, wherein the traffic violation behavior data includes violation images and violation types of motorcycles and electric bicycles, and the traffic violation behavior data group includes violation frame image groups and violation type groups of motorcycles and electric bicycles; performing face and license plate recognition processing on the violation images or violation frame images to obtain face recognition results and license plate recognition results; and judging whether the face recognition result is invalid. The face data and the license plate recognition result are invalid license plate data. If it is judged as no, identity recognition calibration processing is performed according to the face recognition result and / or license plate recognition result, the violator information is obtained and the violator is notified to carry out online or offline education processing according to the violator information; if it is judged as yes, the license plate RFID information group of the corresponding time period is obtained from the license plate RFID collection database according to the currently determined violation time, identity screening processing is performed according to the license plate RFID information group, the violator information is obtained and the violator is notified to carry out online or offline education processing according to the violator information, wherein the license plate RFID information group includes at least one license plate RFID information.

[0005] As an improvement to the above-mentioned scheme, the step of performing identity recognition calibration processing based on the face recognition results and / or license plate recognition results includes: comparing the face recognition results with the face information in the public security database to obtain the identity information of the people on the vehicle, wherein the identity information of the people on the vehicle includes the driver's identity information and / or the passenger's identity information; comparing the license plate recognition results with the license plate registration information in the public security database to obtain the owner's identity information; constructing an identity recognition report based on the identity information of the people on the vehicle and / or the owner's identity information, and determining the information of the violator based on the identity recognition report.

[0006] As an improvement to the above-mentioned scheme, the step of determining the information of the violator based on the identity identification report includes: when the identity information of the person on the vehicle in the identity identification report is successfully matched with the identity information of the vehicle owner or the identity identification report only contains the identity information of the vehicle owner, the information of the violator is determined based on the identity information of the vehicle owner; when the identity information of the person on the vehicle in the identity identification report is not successfully matched with the identity information of the vehicle owner or the identity identification report only contains the identity information of the person on the vehicle, the information of the violator is determined based on the identity information of the person on the vehicle.

[0007] As an improvement to the above scheme, the step of performing identity screening based on the license plate RFID information group also includes: obtaining a vehicle owner identity information group based on the comparison of the license plate RFID information group with the RFID tag data in the public security database; judging whether there is only one vehicle owner identity information in the vehicle owner identity information group; if so, the vehicle owner identity information is the information of the violator; if not, the vehicle owner identity information in the vehicle owner identity information group is manually screened to determine the information of the violator.

[0008] As an improvement to the above-mentioned scheme, the step of conducting online or offline education and processing includes: judging whether the cumulative number of violations of the violator within a preset time period is greater than a preset number; if it is judged as yes, the violator is subjected to offline education and processing; if it is judged as no, the violator is subjected to offline education and processing.

[0009] As an improvement to the above-mentioned solution, the training steps of the traffic violation detection model include: obtaining an image dataset of motorcycles and electric vehicles and performing image preprocessing to obtain a training dataset and a test dataset, wherein the image dataset includes legal driving image data and illegal driving image data; training a multi-task detection model using the training dataset to obtain a traffic violation detection model, wherein the multi-task network architecture in the traffic violation model includes an input layer, a shared convolutional layer, a task-specific layer, a classification layer, and a multi-task loss weighted layer, each of the task-specific layers having a different convolution kernel; testing and evaluating the traffic violation detection model using the test dataset, and optimizing and adjusting the traffic violation detection model based on the test and evaluation results.

[0010] As an improvement to the above scheme, the step of training and testing the traffic violation model using the training data set and the test data set includes: the input layer is used to receive the training data set and input it into multiple driving violation identification tasks; the shared convolution layer is used to extract common feature data in multiple driving violation identification tasks; the task-specific layer is used to extract corresponding high-level feature data through the specific convolution layer in each driving violation identification task based on the common features; the classification layer is used to fuse the feature data and perform predictive classification through the fully connected layer and the softmax layer in each driving violation identification task to obtain preset data; the multi-task loss weighted layer is used to input the corresponding type of preset data and real data into the loss function in each driving violation identification task, and adjust the weights of each loss function through dynamic weight adjustment rules, thereby dynamically adjusting the model parameters to obtain a trained traffic violation detection model.

[0011] As an improvement to the above scheme, the image preprocessing step includes: classifying the image dataset according to the type of violation and labeling the category labels to obtain a labeled image dataset, wherein the labeled image dataset includes violation labeled image data and legal labeled image data; dividing the labeled image dataset into a training dataset and a test dataset according to a preset proportion distribution rule.

[0012] Correspondingly, the present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the above method when executing the computer program.

[0013] Accordingly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0014] The implementation of the present invention has the following beneficial effects:

[0015] The present invention can detect and identify illegal images and illegal driving behaviors from monitoring images or videos, and can use multiple identity recognition methods to determine the information of violators based on the face and vehicle recognition of the illegal images. The detection efficiency and accuracy are high, so that illegal drivers can be educated online or offline, thereby improving the public's awareness of complying with traffic laws and regulations, reducing the incidence of traffic accidents, and reducing the workload of traffic police's manual patrols, thereby improving the work management efficiency of traffic management departments. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of the traffic violation education management method of the present invention;

[0017] Figure 2 It is a flow chart of the training steps of the traffic violation detection model of the present invention. DETAILED DESCRIPTION

[0018] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be described in further detail below with reference to the accompanying drawings.

[0019] like Figure 1 As shown, a specific embodiment of the present invention provides a traffic violation education management method, including:

[0020] S101, obtaining surveillance images or videos of the traffic area and a license plate RFID collection database;

[0021] It should be noted that surveillance cameras located in different traffic zones in various parts of the city can capture surveillance images or videos of the corresponding traffic zones. Accordingly, RFID monitoring devices located in different traffic zones in various parts of the city can monitor RFID vehicles passing through the corresponding traffic zones in real time and obtain their license plate RFID information, thereby constructing a license plate RFID collection database for the corresponding traffic zones. The RFID monitoring device is preferably an RFID reader or a camera with an RFID module to read the RFID tag information of passing vehicles, which includes a unique identification code and license plate information.

[0022] S102. Inputting the surveillance image or video into a traffic violation detection model to identify traffic violation data or data groups, wherein the traffic violation data includes violation images and violation types of motorcycles and electric bicycles, and the traffic violation data groups include violation frame image groups and violation type groups of motorcycles and electric bicycles;

[0023] It should be noted that for surveillance videos, the video can be decomposed and converted into a group of frame images arranged in sequence. By inputting the surveillance image or frame image group into the traffic violation detection model, the violation images and violation types of motorcycles and electric bicycles can be automatically identified, such as overloading, not wearing a safety helmet, driving on the wrong side of the road, and other violation types.

[0024] S103, performing face and license plate recognition processing on the violation image or violation frame image to obtain a face recognition result and a license plate recognition result;

[0025] It should be noted that the face recognition processing of the violation image or the violation frame image is performed using an existing face recognition model to obtain a face recognition result, wherein the face recognition result includes the face recognition results of all people in the vehicle. The license plate recognition processing of the violation image or the violation frame image is performed using an existing license plate recognition model to obtain a license plate recognition result;

[0026] S104: Determine whether the face recognition result is invalid face data and whether the license plate recognition result is invalid license plate data.

[0027] If the judgment is no, identity recognition calibration is performed based on the face recognition results and / or license plate recognition results, the violator's information is obtained, and the violator is notified based on the violator's information for online or offline education;

[0028] Specifically, the steps of performing identity recognition calibration processing on the face recognition result and / or license plate recognition result include:

[0029] Step 1: Compare the face recognition result with the face information in the public security database to obtain the identity information of the people on the vehicle, wherein the identity information of the people on the vehicle includes the identity information of the driver and / or the identity information of the passengers;

[0030] Step 2: Compare the license plate recognition result with the license plate registration information in the public security database to obtain the vehicle owner's identity information;

[0031] Step 3: construct an identity recognition report based on the identity information of the people on the vehicle and / or the identity information of the vehicle owner, and determine the information of the violator based on the identity recognition report.

[0032] It should be noted that when at least one of the facial recognition results and the license plate recognition results is valid data, the driver's identity information and / or passenger's identity information can be obtained by comparing the facial recognition results with the facial information in the public security database; the owner's identity information can be obtained by comparing the license plate recognition results with the license plate registration information in the public security database; an identity recognition report is constructed based on the driver's identity information and / or passenger's identity information and the owner's identity information, so as to facilitate the subsequent identification of the actual violators based on the identity recognition report and the online or offline education and processing of them.

[0033] The step of determining the violator's information based on the identity recognition report includes:

[0034] Step 1: When the identity information of the person on the vehicle in the identity recognition report is successfully matched with the identity information of the vehicle owner, or when the identity recognition report contains only the identity information of the vehicle owner, the information of the violator is determined based on the identity information of the vehicle owner;

[0035] Step 2: When the identity information of the vehicle occupants in the identity recognition report is not successfully matched with the vehicle owner's identity information or the identity recognition report contains only the identity information of the vehicle occupants, the information of the violator is determined based on the identity information of the vehicle occupants.

[0036] It should be noted that the vehicle owner's identity information or the vehicle occupant's identity information includes the corresponding person's ID information and contact information. When the occupant's identity information in the identification report is successfully matched with the vehicle owner's identity information, the vehicle owner can be determined to be the actual violator, and they will be notified for online or offline education and treatment based on their contact information. If the identification report only contains the vehicle owner's identity information, the vehicle owner will be communicated with based on their contact information to determine the final violator, and they will be notified for online or offline education and treatment based on their contact information.

[0037] When the identity information of the occupants in the identity identification report is not successfully matched with the identity information of the vehicle owner or the identity identification report only contains the identity information of the occupants, it is first determined whether the identity information of the driver exists in the identity information of the occupants. When the driver's identity information exists, the driver can be determined to be the actual violator, and he or she will be notified to undergo online or offline education based on the driver's contact information; if the driver's identity information does not exist, the driver's information will be determined based on the passenger's contact information, and he or she will be notified to undergo online or offline education based on the driver's contact information, thereby achieving precise traffic education for the violating drivers, improving the public's awareness of complying with traffic laws and regulations, and reducing the incidence of traffic accidents.

[0038] Among them, according to the contact information of the relevant personnel, they can be notified by text message or phone call for online or offline education and processing. The text message or phone call will inform the violators of the violations they have committed and the potential dangers that the violations may bring, so as to provide targeted education and processing for the personnel.

[0039] If the judgment is yes, the license plate RFID information group of the corresponding time period is obtained from the license plate RFID collection database according to the currently determined time of the violation, and identity screening is performed based on the license plate RFID information group. The information of the violator is obtained and the violator is notified based on the violator information for online or offline education and processing, wherein the license plate RFID information group includes at least one license plate RFID information.

[0040] It should be noted that when both facial recognition and license plate recognition results are invalid, the identity of the offender cannot be obtained through image recognition. In this case, RFID identification is used to determine the offender's information. Based on the time of the currently determined violation, the license plate RFID information group for the corresponding time period is obtained from the license plate RFID collection database. Based on this license plate RFID information group, the corresponding offender's information can be determined, and they can be notified for online or offline education and treatment.

[0041] Specifically, the step of performing identity screening based on the license plate RFID information group further includes:

[0042] Step 1: Compare the license plate RFID information group with the RFID tag data in the public security database to obtain the vehicle owner identity information group;

[0043] Step 2: Determine whether there is only one vehicle owner identity information in the vehicle owner identity information group. If so, the vehicle owner identity information is the information of the violator. If not, manually screen the vehicle owner identity information in the vehicle owner identity information group to determine the information of the violator.

[0044] It should be noted that the license plate RFID information in the license plate RFID information group is compared one by one with the RFID tag data in the public security database to obtain the owner identity information group of the corresponding RFID vehicle personnel. When there is only one owner identity information in the owner identity information group, the owner identity information contact information is communicated with the person to ultimately determine the actual violator and notify the actual violator for online or offline education and processing. When the owner identity information group includes multiple owner identity information, it is necessary for staff to determine the actual vehicle that caused the corresponding violation based on the actual monitoring situation, thereby obtaining the corresponding owner identity information, and communicating with the person based on the owner identity information contact information to ultimately determine the actual violator and notify the actual violator for online or offline education and processing, ensuring that the person who committed the violation is accurately located, and then conducting precise traffic education and processing, improving the violators' awareness of complying with traffic laws and regulations, and reducing the incidence of traffic accidents.

[0045] Furthermore, it is determined whether the cumulative number of violations committed by the violator within a preset time period is greater than a preset number. If so, the violator is given offline education; if not, the violator is given offline education.

[0046] It should be noted that within the preset time period, if the cumulative number of violations committed by the violator is greater than the preset number, the violator will be preliminarily determined to be a habitual offender and will need to be notified to undergo offline education and reform to enhance the violator's safety driving awareness. For example, he or she may need to go to a designated department for offline education and reform activities, otherwise the violating vehicle will be temporarily detained. If the cumulative number of violations committed by the violator is less than the preset number, the violator will be preliminarily determined to be a first-time offender and will need to be notified to undergo online education and reform to improve the driver's safety driving awareness, such as logging into the violation education applet through WeChat to learn the traffic rules for motorcycles and electric automatic vehicles.

[0047] The Violation Education Mini-Program includes both a public-facing and auxiliary-police education mini-program. The public-facing mini-program allows violators to follow online educational videos and share their learning progress on social media. The auxiliary-police education mini-program allows traffic officers to track violators, identify road sections with frequent violations, collect violator information, and review violator education results. This effectively reduces traffic officer workload, saves labor costs, and effectively improves education management efficiency, raises public awareness of traffic regulations, and reduces the incidence of traffic accidents.

[0048] Preferably, the preset time period is 1 year, but is not limited thereto and can be adjusted according to actual needs.

[0049] Preferably, the preset number of times is 3 times, but it is not limited thereto and can be adjusted according to actual needs.

[0050] Furthermore, if Figure 2 As shown, the training steps of the traffic violation detection model include:

[0051] S201, obtaining an image dataset of motorcycles and electric vehicles and performing image preprocessing to obtain a training dataset and a test dataset, wherein the image dataset includes legal driving image data and illegal driving image data;

[0052] Specifically, the image preprocessing step includes:

[0053] Step 1: classify the image dataset according to the type of violation and label the category label to obtain a labeled image dataset, wherein the labeled image dataset includes violation labeled image data and legal labeled image data;

[0054] Step 2: Divide the labeled image dataset into a training dataset and a test dataset according to the preset proportion distribution rule.

[0055] It should be noted that, based on the type of violation, the image dataset can be classified into illegal driving image data and legal driving image data. The illegal driving image data is then subdivided to obtain illegal driving image data of different violation types. Finally, the classified image data is annotated with category labels to obtain a labeled image dataset. In order to train the model and optimize its performance, the labeled image dataset is divided into a training dataset and a test dataset according to a preset ratio allocation rule. The preset ratio allocation rule is an 8:2 ratio between training and testing, that is, the labeled image dataset is divided into a training dataset (80%) and a test dataset (20%) according to a ratio of 80% and 20% respectively. The preset ratio allocation rule is not limited to this and can be adjusted according to actual needs.

[0056] S202. Training a multi-task detection model using the training data set to obtain a traffic violation detection model, wherein the multi-task network architecture in the traffic violation model includes an input layer, a shared convolutional layer, a task-specific layer, a classification layer, and a multi-task loss weighting layer;

[0057] Specifically, the steps of training and testing the traffic violation model using the training data set and the test data set include:

[0058] The input layer is used to receive the training data set and input it into multiple driving violation recognition tasks;

[0059] The shared convolutional layer is used to extract common feature data in multiple driving violation recognition tasks;

[0060] The task-specific layer is used to extract corresponding high-level feature data based on the general features through a specific convolutional layer in each driving violation recognition task;

[0061] It should be noted that branch networks can be designed for different driving violation identification tasks, and different convolution kernels (i.e., feature convolution layers) can be set in different branch networks, including the number and size of convolution kernels. For example, the convolution kernel sizes for identifying overloaded vehicles, identifying vehicles without helmets, and identifying vehicles that occupy the lane can be designed to be (3×3), (1×1), and (5×5), respectively. The branch network for overloaded vehicles can capture the overall characteristics of the vehicle, the branch network for vehicles without helmets can capture local features (such as the head area), and the branch network for vehicles that occupy the lane can extract multi-scale features to better capture the relative position of the vehicle and the road boundary, improving the feature recognition and extraction performance of each task, thereby improving the model detection accuracy.

[0062] The classification layer is used to fuse the feature data through the fully connected layer and the softmax (activation function) layer in each driving violation identification task and perform prediction classification to obtain preset data;

[0063] The multi-task loss weighting layer is used to input the corresponding type of preset data and real data into the loss function in each illegal driving identification task, and adjust the weight of each loss function through dynamic weight adjustment rules, thereby dynamically adjusting the model parameters to obtain a trained traffic violation detection model.

[0064] It should be noted that a corresponding loss function is designed for each illegal driving identification task. For example, for overloaded driving and not wearing a helmet identification tasks, binary cross-entropy loss (Binary Cross-Entropy Loss) is used, and for lane-occupying driving identification tasks, positioning loss (such as IoU loss) and classification loss are integrated. The loss functions of the above-mentioned multiple tasks are combined, and the corresponding weight coefficient is assigned to the loss function of each task to obtain a multi-task loss weighted layer. When used, the model parameters can be dynamically adjusted through the multi-task loss weighted layer to train a traffic violation detection model. Among them, the dynamic weight adjustment rule can preferably be gradient normalization (GradNorm) or adaptive weight distribution method to dynamically adjust the weight coefficient to balance their contributions, ensure that the model can achieve good performance on all tasks, and improve data detection accuracy.

[0065] S203: Testing and evaluating the traffic violation detection model using the test data set, and optimizing and adjusting the traffic violation detection model according to the test and evaluation results.

[0066] It should be noted that after training the traffic violation detection model, the model's performance is evaluated on a test set of violations such as overloading, not wearing a safety helmet, and driving on the wrong lane. The performance of each task is measured using indicators such as accuracy and recall rate. Finally, the model parameters and the weight of the loss function are adjusted based on the evaluation results, thereby optimizing and adjusting the traffic violation detection model to achieve high-accuracy violation task detection.

[0067] Correspondingly, the present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the above method when executing the computer program.

[0068] Accordingly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0069] To sum up, the present invention can detect and identify illegal images and illegal driving behaviors from monitoring images or videos, and can use multiple identity recognition methods to determine the information of violators based on the face and vehicle recognition of the illegal images. The detection efficiency and accuracy are high, thereby providing online or offline education for illegal drivers, improving the public's awareness of complying with traffic laws and regulations, reducing the incidence of traffic accidents, and reducing the workload of traffic police's manual patrols, thereby improving the work management efficiency of traffic management departments.

[0070] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or replace some of the technical features therein with equivalents. Such modifications, changes, or replacements do not deviate from the essence of the corresponding technical solutions within the scope of the technical solutions of the embodiments of the present invention, and should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A traffic violation education and management method, characterized in that: include: Obtain surveillance images or videos of traffic areas, as well as license plate RFID collection database; Inputting the surveillance image or video into a traffic violation detection model to identify traffic violation data or data groups, wherein the traffic violation data includes violation images and violation types of motorcycles and electric bicycles, and the traffic violation data groups include violation frame image groups and violation type groups of motorcycles and electric bicycles; Performing face and license plate recognition processing on the violation image or violation frame image to obtain a face recognition result and a license plate recognition result; Determine whether the face recognition result is invalid face data and whether the license plate recognition result is invalid license plate data, If the judgment is no, identity recognition calibration is performed based on the face recognition results and / or license plate recognition results, the violator's information is obtained, and the violator is notified based on the violator's information for online or offline education; If the judgment is yes, the license plate RFID information group of the corresponding time period is obtained from the license plate RFID collection database according to the currently determined time of the violation, and identity screening is performed based on the license plate RFID information group. The information of the violator is obtained and the violator is notified based on the violator information for online or offline education and processing, wherein the license plate RFID information group includes at least one license plate RFID information.

2. The traffic violation education and management method according to claim 1, characterized in that: The step of performing identity recognition calibration processing based on the face recognition result and / or the license plate recognition result includes: Comparing the facial recognition result with facial information in a public security database to obtain identity information of the people on board the vehicle, wherein the identity information of the people on board the vehicle includes the identity information of the driver and / or the identity information of the passengers; Comparing the license plate recognition result with the license plate registration information in the public security database to obtain the vehicle owner's identity information; An identity recognition report is constructed based on the identity information of the persons on the vehicle and / or the identity information of the vehicle owner, and information on the violating persons is determined based on the identity recognition report.

3. The traffic violation education and management method according to claim 2, characterized in that: The step of determining the violator's information based on the identity recognition report includes: When the identity information of the person on the vehicle in the identity recognition report is successfully matched with the identity information of the vehicle owner or the identity recognition report contains only the identity information of the vehicle owner, the information of the violator is determined based on the identity information of the vehicle owner; When the identity information of the person on the vehicle in the identity recognition report is not successfully matched with the identity information of the vehicle owner or the identity recognition report contains only the identity information of the person on the vehicle, the information of the violator is determined based on the identity information of the person on the vehicle.

4. The traffic violation education and management method according to claim 1, characterized in that: The step of performing identity screening based on the license plate RFID information group further includes: Comparing the license plate RFID information group with the RFID tag data in the public security database to obtain the vehicle owner identity information group; Determine whether there is only one vehicle owner identity information in the vehicle owner identity information group. If so, the vehicle owner identity information is the information of the violator. If not, manually screen the vehicle owner identity information in the vehicle owner identity information group to determine the information of the violator.

5. The traffic violation education and management method according to claim 1, characterized in that: The steps of conducting online or offline education processing include: Determine whether the cumulative number of violations committed by the violator within the preset time period is greater than the preset number. If so, provide offline education to the violator. If not, provide offline education to the violator.

6. The traffic violation education and management method according to claim 1, characterized in that: The training steps of the traffic violation detection model include: Acquire image datasets of motorcycles and electric vehicles and perform image preprocessing to obtain training datasets and test datasets, wherein the image datasets include legal driving image data and illegal driving image data; Training a multi-task detection model using the training dataset to obtain a traffic violation detection model, wherein the multi-task network architecture in the traffic violation model includes an input layer, a shared convolutional layer, a task-specific layer, a classification layer, and a multi-task loss weighted layer, wherein each task-specific layer has a different convolution kernel; The traffic violation detection model is tested and evaluated using the test data set, and the traffic violation detection model is optimized and adjusted based on the test and evaluation results.

7. The traffic violation education and management method according to claim 6, characterized in that: The steps of training and testing the traffic violation model using the training data set and the test data set include: The input layer is used to receive the training data set and input it into multiple driving violation recognition tasks; The shared convolutional layer is used to extract common feature data in multiple driving violation recognition tasks; The task-specific layer is used to extract corresponding high-level feature data based on the general features through a specific convolutional layer in each driving violation recognition task; The classification layer is used to fuse the feature data through the fully connected layer and the softmax layer in each driving violation identification task and perform prediction classification to obtain preset data; The multi-task loss weighting layer is used to input the corresponding type of preset data and real data into the loss function in each illegal driving identification task, and adjust the weight of each loss function through dynamic weight adjustment rules, thereby dynamically adjusting the model parameters to obtain a trained traffic violation detection model.

8. The traffic violation education and management method according to claim 6, characterized in that: The image preprocessing step includes: Classifying the image dataset according to the type of violation and labeling the category labels to obtain a labeled image dataset, wherein the labeled image dataset includes violation labeled image data and legal labeled image data; The labeled image dataset is divided into a training dataset and a test dataset according to the preset proportion distribution rules.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images