Mining rubber-tyred vehicle driver behavior monitoring method and system based on deep learning

Abstract

The present invention discloses a method and system for monitoring the driver behavior of a rubber-tyred mining vehicle based on deep learning, and relates to the technical field of underground mining equipment. A trained behavior detection model is deployed in an edge device, and an image containing a driver is input into the trained behavior detection model. A TPU (tensor processing unit) is used to accelerate reasoning to obtain a target detection result. The result output by the behavior detection model includes the coordinates, confidence and category information of the target object. Based on the output result of the model, it is determined whether the driver has any illegal behavior, and the illegal behavior includes driver fatigue, driver distraction, driver not wearing a safety helmet, driver not wearing a seat belt, driver using a mobile phone and driver smoking. The method and system for monitoring the driver behavior of a rubber-tyred mining vehicle can monitor the driver's status in real time and efficiently identify illegal behavior.

Description

Technical Field

[0001] The present invention relates to the technical field of underground mine equipment, and in particular to a method and system for monitoring the driver behavior of a rubber-tyred mining vehicle based on deep learning. Background Art

[0002] In mining operations, rubber-tyred trackless vehicles, as a key means of transportation, are widely used in the handling of materials such as ore and coal. However, despite its important role in improving transportation efficiency and reducing labor costs, safety has always been an issue of urgent concern. Due to factors such as driver operating errors, improper driving behavior, and the complexity of the surrounding road environment, accidents involving rubber-tyred trackless vehicles occur from time to time, which not only causes economic and property losses, but also may pose a serious threat to personnel safety. In actual work, behaviors such as driver distraction, fatigue driving, and misjudgment of vehicle characteristics are often the root causes of accidents. In addition, the mining environment is complex and changeable, including poor road conditions, limited visibility, and narrow driving paths, which further increase the risk of driving. Summary of the invention

[0003] Based on the technical problems existing in the background technology, the present invention proposes a method and system for monitoring the driver behavior of a rubber-tyred mining vehicle based on deep learning, which can monitor the driver's status in real time and efficiently identify illegal behaviors.

[0004] The method for monitoring driver behavior of a mining rubber-tyred vehicle based on deep learning proposed in the present invention inputs an image containing the driver into a trained behavior detection model to output whether the driver has any illegal behavior, wherein the illegal behavior includes driver fatigue, driver distraction, driver not wearing a helmet, driver not wearing a seat belt, driver using a mobile phone, and driver smoking;

[0005] The behavior detection model is used to analyze the driver's eye and head features. If the model outputs a fatigue label and the duration exceeds the set threshold, the driver is judged to be in a fatigued state, i.e., the driver is fatigued;

[0006] The behavior detection model is used to analyze the driver's head direction. When the driver is detected to be deviating from the label in front and continues to be distracted for more than a set time, it is determined to be a distracted state, that is, the driver is distracted;

[0007] The behavior detection model is used to analyze whether the input image contains a helmet label. If the helmet label does not appear on the driver's head, it is determined that the driver is not wearing a helmet.

[0008] The behavior detection model is used to analyze whether the input image contains a seat belt label. If there is no seat belt label, it is determined that the driver is not wearing a seat belt.

[0009] The behavior detection model is used to analyze the presence and duration of the mobile phone tag. If the model outputs the mobile phone tag and the duration exceeds the set threshold, the mobile phone use behavior is determined, that is, the driver uses the mobile phone;

[0010] The behavior detection model is used to analyze whether the input image contains a smoking label. If there is a smoking label, it is determined that there is smoking behavior, that is, the driver is smoking.

[0011] Furthermore, in driver fatigue detection, if a yawn tag is detected and the number of yawns exceeds a preset threshold within a specific time, it is also determined to be a fatigue state.

[0012] Furthermore, the training process of the behavior detection model is as follows:

[0013] Obtain and filter the driver's driving behavior video, use the Python frame extraction script to convert the filtered video into image data, clean and annotate the image data, and divide the annotated image data into training set, validation set, and test set;

[0014] Build a behavior detection model with Darknet as the backbone network;

[0015] The training set is input into the behavior detection model for training to obtain the behavior detection model after preliminary training. The behavior detection model is verified and tested using the verification set and the test set, and the training set is combined with cyclic training until the behavior detection model converges.

[0016] Furthermore, during the video acquisition process, the camera is installed at the front end of the steering wheel of the rubber-wheeled vehicle to fully capture the driver's position, and the monitoring area of ​​the camera is only for the main driving position;

[0017] During the execution process after the behavior detection model training is completed, it is first determined whether the center point of the detected target frame is within the preset detection area;

[0018] If yes, continue to determine whether the face is successfully detected;

[0019] If the face is successfully detected, other objects will be identified to determine whether the driver has violated the regulations.

[0020] If the center point of the detected target frame is not within the preset detection area, or the face is not successfully detected, the frame image is skipped and new image data is continuously acquired for processing.

[0021] Furthermore, it is determined whether the detected target frame center point is within the preset detection area, specifically:

[0022] The preset detection area is represented as (X1, Y1, X2, Y2), that is, the coordinate point of the upper left corner in the image is (X1, Y1) and the coordinate point of the lower right corner is (X2, Y2);

[0023] The center point of the detected target box is represented as Among them, the coordinate point of the upper left corner of the image is detected to be (x1, y1) and the coordinate point of the lower right corner is (x2, y2);

[0024] when as well as It is determined that the center point of the detected target frame is located in the preset detection area.

[0025] Furthermore, the behavior detection model is equipped with a camera occlusion detection function, which determines whether the camera is blocked by calculating the pixel variance of the acquired image. When the pixel values ​​in the image tend to be consistent, the camera is determined to be in an occluded state. At this time, it is recorded and an alarm is issued, and the message is uploaded to the monitoring platform.

[0026] A mining rubber-tyred vehicle driver behavior monitoring system based on deep learning, comprising a data acquisition module, a first data transmission module, a data processing module, a data storage module, a second data transmission module and a data display module;

[0027] The data acquisition module includes a plurality of network cameras to acquire video including the driver;

[0028] The input of the first data transmission module is connected to one or more network cameras, and the output is connected to the edge device of the data processing module to transmit the video to the edge device;

[0029] The data processing module includes multiple edge devices, each of which is equipped with a trained behavior detection model to output whether the driver has violated the regulations.

[0030] The input of the second data transmission module is connected to the output of all edge devices, the input is connected to the input of the data display module, and the image screen with illegal behavior is transmitted to the data storage module;

[0031] The data display module displays and warns whether the driver has any violations.

[0032] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the driver behavior monitoring method as described above when executing the computer program.

[0033] A computer-readable storage medium stores a plurality of classification programs, wherein the plurality of classification programs are used to be called by a processor and execute the driver behavior monitoring method as described above.

[0034] The advantages of the deep learning-based mining rubber-wheeled vehicle driver behavior monitoring method and system provided by the present invention are: using a convolutional network with Darknet as the backbone, the system can accurately capture a series of detailed behaviors such as driver fatigue, distraction, not wearing a helmet, not wearing a seat belt, using a mobile phone and smoking, ensuring real-time monitoring of the driver's status and efficient identification of violations; in the hardware transplantation part, the system uses the domestic embedded device SE7 of Suanneng as the core edge processor, and collects driver status data in real time through a network camera and transmits it to the edge device for processing. To accelerate video decoding, the Suanneng API sail.Decoder is used for efficient decoding. The optimal torch model obtained by deep convolutional network training is converted to ONNX format, and after pruning and quantization, the ONNX model is converted to bmodel format suitable for SE7 edge devices. Finally, the obtained bmodel format model is deployed in the edge device, and the network camera is used to realize the monitoring of the driver's real-time behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the process of the present invention;

[0036] Figure 2 Generate a flow chart for the behavior detection model;

[0037] Figure 3 This is the architecture diagram of the driver behavior monitoring system. DETAILED DESCRIPTION

[0038] Below, the technical solution of the present invention is described in detail through specific embodiments. Many specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific implementation disclosed below.

[0039] Rubber-tyred vehicles run underground most of the time. If the on-board data is transmitted back to the ground for processing and then the alarm information is fed back, it will take a lot of time and real-time response cannot be achieved. Therefore, it is chosen to deploy edge computing devices on the vehicle side to process the on-site data locally, so that real-time alarms can be achieved when violations occur, thereby ensuring the rapidity of data analysis and response. Edge computing devices significantly reduce data transmission delays and can perform computing tasks locally without relying on central servers. This not only saves time, but also makes more efficient use of computing resources. In addition, its energy consumption is lower than that of traditional computers, because local processing reduces the need for frequent connections to central servers, thereby effectively reducing energy consumption and overall operating costs, and improving the utilization efficiency of computing resources. Therefore, this embodiment is provided to solve the real-time detection of drivers.

[0040] like Figures 1 to 3 As shown, the deep learning-based mining rubber-tyred vehicle driver behavior monitoring method proposed in the present invention inputs an image containing the driver into a trained behavior detection model to output whether the driver has any violations, and the violations include driver fatigue, driver distraction, driver not wearing a safety helmet, driver not wearing a seat belt, driver using a mobile phone and driver smoking.

[0041] In this embodiment, a deep learning algorithm with Darknet as the backbone network is used to collect the driver's driving behavior data in real time by installing a camera in the cab of a rubber-wheeled vehicle. After data cleaning and frame processing, a driver behavior detection data set is formed. The processed data set is trained using a deep learning algorithm to obtain a preliminary driver behavior detection model. Subsequently, the model is tested using a validation set, and the data set is continuously adjusted based on the test results, and finally an optimal driver behavior detection model is trained. Figure 2 As shown, the training process of the behavior detection model is as follows (a1) to (a3):

[0042] (a1) Obtain and filter the driver's driving behavior video, use the python frame extraction script to convert the filtered video into image data, clean and annotate the image data, divide the annotated image data into training set, validation set and test set according to 8:1:1, set the number of training rounds (epochs) to 280, the batch size (batch size) to 128, and the learning rate (learning rate) to 0.01. The training equipment uses dual NVIDIA GeForce RTX 3090 graphics cards to improve computing efficiency and training speed;

[0043] (a2) Build a behavior detection model using Darknet as the backbone network;

[0044] (a3) Inputting the training set into the behavior detection model for training, obtaining the behavior detection model after preliminary training, using the validation set and the test set to verify and test the behavior detection model, and combining the cyclic training of the training set until the behavior detection model converges.

[0045] The trained behavior detection model is then used to determine whether the driver has violated the rules. In this embodiment, in order to prevent other passengers from affecting the detection results, the system limits the detection area and only monitors the main driver's seat. During driver detection, whether the driver wears a seat belt and a helmet is determined by detecting whether there are corresponding labels in the specified area; whether the driver violates the rules by smoking or using a mobile phone, the system identifies whether there are mobile phone and cigarette labels in the detection area; to determine whether the driver is distracted, the system determines by detecting long periods of head turning; in addition, the system is also equipped with a camera occlusion detection function, which determines whether the camera is blocked by calculating the pixel variance of the image. When the pixel values ​​in the image tend to be consistent, it is determined that the camera is in an obstructed state. At this time, the driver detection is not entered, and the message is recorded and an alarm is issued, and the message is uploaded to the monitoring platform. The specific process of driver detection is (b1) to (b7):

[0046] (b1) After the system is started, it first determines whether the center point of the detected target frame is within the preset detection area (detection_area), and then determines whether the face is successfully detected. When the face is detected normally, the system continues to analyze other targets for judgment to determine whether there is any violation. If the center point of the detected target frame is not within the preset detection area, or the face is not successfully detected, the frame image is skipped and new image data is acquired for processing.

[0047] To determine whether the center point of the detected target frame is within the preset detection area, specifically: the preset detection area is represented by (X1, Y1, X2, Y2), that is, the coordinate point of the upper left corner in the image is (X1, Y1) and the coordinate point of the lower right corner is (X2, Y2); the center point of the detected target frame is represented by Among them, the coordinate point of the upper left corner of the image is detected to be (x1, y1) and the coordinate point of the lower right corner is (x2, y2); when as well as It is determined that the center point of the detected target frame is located in the preset detection area.

[0048] (b2) By acquiring the driver's video and slicing it, the behavior detection model is used to analyze the driver's eye and head features. If the model outputs a fatigue label and the duration exceeds the set threshold, the driver is judged to be in a fatigue state, i.e., the driver is tired; if a yawn label is detected and the number of yawns exceeds the preset threshold within a specific time, it is also judged to be a fatigue state.

[0049] (b3) The driver's video is obtained and sliced, and the driver's head direction is analyzed using a behavior detection model. When a label deviating from the front is detected and lasts for more than a set time, it is determined to be a distracted state, that is, the driver is distracted;

[0050] (b4) By acquiring the driver video and slicing it, the behavior detection model is used to analyze whether the input image contains a helmet label. If the helmet label does not appear on the driver's head, it is determined that the driver is not wearing a helmet.

[0051] (b5) The driver video is obtained and sliced, and the behavior detection model is used to analyze whether the input image contains a seat belt label. If there is no seat belt label, it is determined that the driver is not wearing a seat belt, that is, the driver is not wearing a seat belt;

[0052] (b6) By acquiring the driver video and slicing it, the behavior detection model is used to analyze the presence and duration of the mobile phone tag. If the model outputs the mobile phone tag and the duration exceeds the set threshold, the mobile phone use behavior is determined, that is, the driver uses the mobile phone;

[0053] (b7) By acquiring the driver video and slicing it, the behavior detection model is used to analyze whether the input image contains a smoking label. If there is a smoking label, it is determined that there is smoking behavior, that is, the driver is smoking.

[0054] This embodiment uses a deep learning algorithm to continuously analyze the driver's video in multiple frames to identify the driver's facial expressions and behavioral actions. Using a convolutional network with Darknet as the backbone, the system can accurately capture a series of detailed behaviors such as driver fatigue, distraction, not wearing a helmet, not wearing a seat belt, using a mobile phone, and smoking, ensuring real-time monitoring of the driver's status and efficient identification of violations.

[0055] In addition, corresponding detection processes are designed according to different violations, and accurate identification is achieved step by step through processes such as fatigue detection, distraction recognition, helmet and seat belt detection, mobile phone use and smoking discrimination. The independence and pertinence of each detection process improves the system's generalization ability in complex application scenarios, enables the system to adapt to diverse driving behavior characteristics, effectively reduces false alarm rates, and improves reliability and practicality in actual applications.

[0056] In this embodiment, the trained behavior detection model is deployed in the edge intelligent computing module to determine the driver's violation behavior.

[0057] like Figure 3As shown, the deep learning-based driver behavior monitoring system for rubber-tyred mining vehicles includes an edge intelligent computing module, a ground monitoring platform, a voice alarm module, and a data storage module, wherein the edge intelligent computing module includes a data acquisition module, a first data transmission module, a data processing module, a data storage module, a second data transmission module, and a data display module.

[0058] The data acquisition module includes a plurality of network cameras to acquire video including the driver;

[0059] The input of the first data transmission module is connected to one or more network cameras, and the output is connected to the edge device of the data processing module to transmit the video to the edge device;

[0060] The data processing module includes multiple edge devices, each of which is equipped with a trained behavior detection model to output whether the driver has violated the rules. Specifically, the processed image data will be fed into the behavior detection model frame by frame, and the TPU (tensor processing unit) will be used to accelerate the reasoning to obtain the target detection results. The output of the behavior detection model includes the coordinates, confidence and category information of the target object. Based on the output of the model, the system will determine whether the driver has violated the rules. If the judgment result shows that the driver has violated the rules, the system will use the voice alarm function to remind the driver of the corresponding violation, and will automatically capture the screen of the violation for archiving;

[0061] The input of the second data transmission module is connected to the output of all edge devices, the input is connected to the input of the data display module, and the image screen with illegal behavior is transmitted to the data storage module;

[0062] The data display module displays whether the driver has violated the rules. If there is a violation, the screen and platform alarm information will be transmitted to the ground monitoring platform through the network switch to ensure that relevant information can be fed back to the monitoring personnel in real time. At the same time, the video data will also be stored in the local server for subsequent review and analysis.

[0063] The edge intelligent computing module includes a network camera for data collection, a network switch for data transmission, and a domestic edge device SE7 for data processing. The SE7 edge device is equipped with two 10 / 100 / 1000Mbps adaptive network ports and a wireless network card module. When the host computer and the SE7 device are in the same network segment, the host computer is allowed to interact and communicate with the edge device through the IP address. The host computer can control and operate the edge device to achieve effective management and configuration of the system.

[0064] Each network camera is installed in the cab of the rubber-wheeled vehicle to obtain the real-time status data of the driver. Multiple network cameras are connected to the same edge intelligent computing device through a network switch. The edge intelligent computing module uses the image processing module and deep learning model to perform AI detection and analysis on the real-time status of the driver. The processed video data is transmitted to the ground control module through the network switch for display. At the same time, the processed video data is stored in the server. When a violation occurs, the status information will be sent to the ground monitoring platform to generate an alarm message and send an alarm message to the voice alarm module. At the same time, the corresponding violation will be recorded by screenshot.

[0065] The ground monitoring platform is used to receive the results processed by the edge intelligent computing module and display them on the large screen in the ground control room. The large screen can display more than 8 video streams at the same time, and will issue warning reminders when the driver has abnormal behavior;

[0066] The voice alarm module is placed in the cab of the rubber-wheeled vehicle, and receives the alarm signal from the edge intelligent computing module to generate a corresponding voice alarm to remind the driver to drive properly;

[0067] The data storage module refers to a storage server or cloud server used to store the driver status video data processed by the edge intelligent computing module.

[0068] In this embodiment, after data processing, when a driver is detected to have violated the rules, the driver will be promptly reminded through voice broadcast. At the same time, the relevant violation records will be stored in the edge device, and ground personnel can access and view the data through the IP address of the device. This mechanism ensures timely feedback of violations and convenient subsequent management;

[0069] In addition, this embodiment adopts a modular design, and each module works closely together to form a smooth monitoring process from data acquisition to data processing to result display. The network camera is responsible for real-time acquisition of driver images, the edge computing module performs data cleaning, decoding, and analysis, the voice alarm module promptly feedbacks violation information, the ground monitoring platform displays alarm information and real-time images, and the data storage module retains the detection data for subsequent reference. Multi-module collaboration enables the system to be efficient, real-time, and highly accurate, effectively improving the level of intelligent management of driving safety of rubber-wheeled vehicles in mines.

[0070] In order to ensure the stable use of edge devices, consider using domestic edge devices. The computing equipment is equipped with self-developed TPU processors, such as BM1684 and BM1684x, and also provides a variety of modules, such as SM5 and SM7, as well as edge device models SE5 and SE7. These domestic TPU processors can support a variety of mainstream deep learning frameworks, including Caffe, TensorFlow, PyTorch, ONNX and PaddlePaddle, ensuring efficient performance and compatibility in a variety of application scenarios, thereby further promoting the autonomous and controllable development of AI technology.

[0071] Therefore, in the hardware transplantation part of this embodiment, the system uses the domestic embedded device SE7 of Suanneng as the core edge processor, collects the driver status data in real time through the network camera and transmits it to the edge device for processing. In order to accelerate video decoding, the Suanneng API sail.Decoder is used for efficient decoding. The optimal torch model obtained by deep convolutional network training is converted to ONNX format, and after pruning and quantization, the ONNX model is converted to the bmodel format suitable for SE7 edge devices. Finally, the obtained bmodel format model is deployed in the edge device, and the real-time behavior of the driver is monitored in conjunction with the network camera. Based on the output results of the model, the system will automatically determine whether the driver has violated the rules. If a violation is detected, the system will issue a reminder in time through the voice alarm function, and automatically capture the screen of the violation for archiving. In addition, the monitoring screen and platform alarm information will be transmitted to the ground monitoring platform through the network switch to ensure that the relevant information can be fed back to the monitoring personnel in real time. At the same time, the video data will also be stored in the local server for subsequent reference and analysis.

[0072] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for monitoring driver behavior of a mining rubber-tyred vehicle based on deep learning, characterized in that: Inputting the image containing the driver into the trained behavior detection model to output whether the driver has any violation, including driver fatigue, driver distraction, driver not wearing a helmet, driver not wearing a seat belt, driver using a mobile phone, and driver smoking; The behavior detection model is used to analyze the driver's eye and head features. If the model outputs a fatigue label and the duration exceeds the set threshold, the driver is judged to be in a fatigued state, i.e., the driver is fatigued; The behavior detection model is used to analyze the driver's head direction. When the driver is detected to be deviating from the label in front and continues to be distracted for more than a set time, it is determined to be a distracted state, that is, the driver is distracted; The behavior detection model is used to analyze whether the input image contains a helmet label. If the helmet label does not appear on the driver's head, it is determined that the driver is not wearing a helmet. The behavior detection model is used to analyze whether the input image contains a seat belt label. If there is no seat belt label, it is determined that the driver is not wearing a seat belt. The behavior detection model is used to analyze the presence and duration of the mobile phone tag. If the model outputs the mobile phone tag and the duration exceeds the set threshold, the mobile phone use behavior is determined, that is, the driver uses the mobile phone; The behavior detection model is used to analyze whether the input image contains a smoking label. If there is a smoking label, it is determined that there is smoking behavior, that is, the driver is smoking.

2. The method for monitoring driver behavior of a mining rubber-tyred vehicle based on deep learning according to claim 1 is characterized in that: In driver fatigue detection, if a yawn tag is detected and the number of yawns exceeds a preset threshold within a specific time, it is also determined to be a fatigue state.

3. The method for monitoring driver behavior of a mining rubber-tyred vehicle based on deep learning according to claim 1 is characterized in that: The training process of the behavior detection model is as follows: Obtain and filter the driver's driving behavior video, use the Python frame extraction script to convert the filtered video into image data, clean and annotate the image data, and divide the annotated image data into training set, validation set, and test set; Build a behavior detection model with Darknet as the backbone network; The training set is input into the behavior detection model for training to obtain the behavior detection model after preliminary training. The behavior detection model is verified and tested using the verification set and the test set, and the training set is combined with cyclic training until the behavior detection model converges.

4. The method for monitoring driver behavior of a mining rubber-tyred vehicle based on deep learning according to claim 1 is characterized in that: During the video acquisition process, the camera is installed at the front of the steering wheel of the rubber-wheeled vehicle to fully capture the driver's position. The camera's monitoring area is only for the main driving position; During the execution process after the behavior detection model training is completed, it is first determined whether the center point of the detected target frame is within the preset detection area; If yes, continue to determine whether the face is successfully detected; If the face is successfully detected, other objects will be identified to determine whether the driver has violated the regulations. If the center point of the detected target frame is not within the preset detection area, or the face is not successfully detected, the frame image is skipped and new image data is continuously acquired for processing.

5. The method for monitoring driver behavior of a mining rubber-tyred vehicle based on deep learning according to claim 4 is characterized in that: Determine whether the center point of the detected target frame is within the preset detection area, specifically: The preset detection area is represented as (X1, Y1, X2, Y2), that is, the coordinate point of the upper left corner in the image is (X1, Y1) and the coordinate point of the lower right corner is (X2, Y2); The center point of the detected target box is represented as Among them, the coordinate point of the upper left corner of the image is detected to be (x1, y1) and the coordinate point of the lower right corner is (x2, y2); when as well as It is determined that the center point of the detected target frame is located in the preset detection area.

6. The method for monitoring driver behavior of a mining rubber-tyred vehicle based on deep learning according to claim 4 is characterized in that: The behavior detection model is equipped with a camera occlusion detection function, which determines whether the camera is blocked by calculating the pixel variance of the acquired image. When the pixel values ​​in the image tend to be consistent, the camera is determined to be in an occluded state. At this time, it is recorded and an alarm is issued, and the message is uploaded to the monitoring platform.

7. The mining rubber-tyred vehicle driver behavior monitoring system based on deep learning is characterized by: It includes a data acquisition module, a first data transmission module, a data processing module, a data storage module, a second data transmission module and a data display module; The data acquisition module includes a plurality of network cameras to acquire video including the driver; The input of the first data transmission module is connected to one or more network cameras, and the output is connected to the edge device of the data processing module to transmit the video to the edge device; The data processing module includes multiple edge devices, each of which is equipped with a trained behavior detection model to output whether the driver has violated the regulations. The input of the second data transmission module is connected to the output of all edge devices, the input is connected to the input of the data display module, and the image screen with illegal behavior is transmitted to the data storage module; The data display module displays and warns whether the driver has any violations.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the driver behavior monitoring method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of classification programs, and the plurality of classification programs are used to be called by a processor and execute the driver behavior monitoring method as described in any one of claims 1-6.

Images