A road condition monitoring system based on video image analysis

Abstract

The present invention relates to a road condition monitoring system based on video image analysis, belonging to the technical field of road monitoring. It measures the speed of target freight vehicles on key sections of the fast lane; when the vehicle speed is lower than a preset speed threshold, it captures the scene to obtain captured video images; it analyzes the traffic conditions at the scene through the captured video images; according to the traffic conditions, it determines whether the target freight vehicle has committed a traffic violation. The present invention solves the problem in the prior art that it is impossible to combine the actual traffic conditions to judge the impact caused by occupying the lane, resulting in possible misjudgment of traffic violations. It identifies the vehicle distribution information behind the target freight vehicle on the fast road, and uses the Bayesian network model to predict the probability of current traffic congestion according to the vehicle distribution information. The present invention improves the accuracy of traffic condition analysis by using the Bayesian network model and provides a reference for traffic decision-making.

Description

Technical Field

[0001] The present invention belongs to the technical field of road monitoring and relates to a road condition monitoring system based on video image analysis. Background Art

[0002] Since trucks are generally large in size and relatively heavy in load, such vehicles tend to travel at a slow speed during driving. If they occupy the overtaking lane for a long time, it will pose a significant safety hazard to other traffic participants. Because in this case, other vehicles cannot overtake from the left and can only choose to overtake from the right. This driving method greatly increases the risk of multi-vehicle accidents such as rear-end collisions and scratches.

[0003] Currently, local traffic police mainly rely on public reports and some set checkpoints for capturing images. However, this method of capturing images at checkpoints can only determine whether the target vehicle occupies the lane and cannot judge the impact of occupying the lane on the actual traffic situation, resulting in the problem of possible misjudgment of traffic violations. Summary of the Invention

[0004] The present invention provides a road condition monitoring system based on video image analysis, aiming to analyze the road condition of key sections by using machine learning methods and solve the problem of misjudgment of traffic violations in the prior art by quantifying the impact of target truck vehicles occupying the lane on the traffic situation.

[0005] The object of the present invention can be achieved by the following technical solutions:

[0006] The present application provides a road condition monitoring system based on video image analysis, including a speed measurement module, a capturing module, an analysis module and a decision-making module. The speed measurement module, the capturing module, the analysis module and the decision-making module are communicatively connected, wherein:

[0007] The speed measurement module is used to measure the speed of target truck vehicles on key sections of the fast lane;

[0008] The capturing module captures the scene to obtain captured video images when the vehicle speed is lower than a preset speed threshold;

[0009] The analysis module analyzes the traffic situation at the scene through the captured video images;

[0010] The decision-making module judges whether there is a traffic violation for the target truck vehicle according to the traffic situation.

[0011] Further, the step of measuring the speed of vehicles on key sections of the fast lane includes the following steps:

[0012] S1. Data collection: Install high-definition cameras at key sections, with the cameras covering the entire lane from the start detection line to the end detection line.

[0013] S2. Object detection: Use the YOLO algorithm to process the collected video stream for detecting and identifying target freight vehicles.

[0014] S3. Speed calculation: After detecting a vehicle, analyze the change in the vehicle's position in consecutive frames to calculate the current driving speed.

[0015] Furthermore, the key sections include main roads, traffic intersections, expressways and highway entrances and exits, congestion hotspots, mountain roads and slope areas, as well as construction or obstacle areas.

[0016] Furthermore, the process of using the YOLO algorithm to process the collected video stream for detecting and identifying vehicles includes the following steps:

[0017] S21. Data preparation: Collect image or video data containing vehicles, use annotation tools to annotate the data, draw the bounding boxes of the vehicles, and label each box with the corresponding label to generate the annotation files required for training.

[0018] S22. Data preprocessing: Convert the annotated data into text file format.

[0019] S23. Environment configuration: Configure the training environment and install deep learning frameworks such as TensorFlow or PyTorch.

[0020] S24. Model training: Set the configuration items for training, which include network structure, input image size, learning rate, and batch size, and then run the training script, monitoring the loss value and performance metrics to ensure that the model gradually converges and reaches the expected accuracy.

[0021] S25. Model evaluation: After completing training, use the test set to evaluate the model and judge the performance of the model by calculating metrics such as mAP, Precision, and Recall.

[0022] Furthermore, the captured video images are taken simultaneously by high-position and low-position cameras; among them, the high-position camera is used to identify the traffic flow and vehicle distribution on the road, and the low-position camera is used to identify vehicle details; the vehicle details include vehicle license plates and vehicle types.

[0023] Furthermore, the analysis of the on-site traffic conditions through the captured video images includes the following steps:

[0024] T1. Identify the vehicle distribution information behind the target truck vehicle on the expressway. The vehicle distribution information includes the number of vehicles, vehicle speeds, and corresponding vehicle behaviors. The vehicle behaviors include acceleration, deceleration, lane change, and overtaking.

[0025] T2. Predict the probability of current traffic congestion based on the vehicle distribution information.

[0026] Further, in step T2, the predicting the probability of current traffic congestion based on the vehicle distribution information includes the following steps:

[0027] T21. Collect traffic condition data of trucks occupying the express lane in the historical period. The traffic condition data includes vehicle distribution information and traffic congestion information. The traffic congestion information is divided into congestion and non-congestion.

[0028] T22. Determine the Bayesian network structure with the vehicle distribution information as the parent node and the traffic congestion information as the child node.

[0029] T23. Use maximum likelihood estimation to determine the conditional probability distribution between the parent node and the child node parameters, and complete the training of the Bayesian network model.

[0030] T24. Input the current vehicle distribution into the trained Bayesian network model to predict the probability of current traffic congestion.

[0031] Further, according to the traffic condition, determine whether the target truck vehicle has traffic violations. Specifically, when the probability of current traffic congestion exceeds a preset first congestion threshold, determine whether the target truck vehicle has traffic violations.

[0032] Further, when the decision module determines that the target truck vehicle has traffic violations, it reminds the target truck vehicle not to occupy the express lane through the traffic indicator light, and retains the captured video image as the basis for violations.

[0033] Further, when the probability of current traffic congestion exceeds a preset second congestion threshold, automatically alarm the traffic management department to request handling of the traffic congestion.

[0034] Advantages of the present invention:

[0035] (1) Measure the speed of the target truck vehicle on the key section of the express lane. When the vehicle speed is lower than the preset speed threshold, capture the scene to obtain the captured video image. Analyze the traffic condition of the scene through the captured video image. According to the traffic condition, determine whether the target truck vehicle has traffic violations. The present invention solves the problem that the prior art cannot combine the actual traffic condition to judge the impact of occupying the lane, resulting in possible misjudgment of traffic violations.

[0036] (2) Identify the vehicle distribution information behind the target truck vehicle on the expressway, and use the Bayesian network model to predict the probability of current traffic congestion according to the vehicle distribution information. The present invention uses the Bayesian network model to improve the accuracy of traffic situation analysis and provides a reference for traffic decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] For the convenience of those skilled in the art to understand, the present invention will be further described below with reference to the accompanying drawings.

[0038] Figure 1 It is a structural diagram of a road condition monitoring system based on video image analysis in the present invention.

[0039] Figure 2 It is a step flowchart of the test module in an embodiment of the present invention.

[0040] Figure 3 It is a flowchart for the present invention to predict the probability of current traffic congestion according to the vehicle distribution information. DETAILED DESCRIPTION OF THE INVENTION

[0041] To further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following will describe in detail the specific implementation manner, structure, features and effects of the present invention with reference to the accompanying drawings and preferred embodiments.

[0042] Please refer to Figures 1 - 3 , this application provides a road condition monitoring system based on video image analysis, including a speed measurement module, a capture module, an analysis module and a decision module. The speed measurement module, the capture module, the analysis module and the decision module are communicatively connected, wherein:

[0043] The speed measurement module is used to measure the speed of the target truck vehicle on the key section of the express lane;

[0044] In this embodiment, the speed measurement module can timely detect and alarm the situation of slow vehicles occupying lanes by monitoring the vehicle driving speed in real time, thereby improving the traffic efficiency of the road. The system uses a high-definition camera to capture vehicle images in a specific section and performs object detection through deep learning algorithms to ensure the accuracy of speed calculation. The system not only records data such as the speed, time, and location of vehicles, but also supports querying historical data, providing data support for traffic management and planning. The dynamically adjustable slow speed threshold can vary flexibly according to the actual traffic flow situation to ensure the effectiveness of monitoring. In addition, the speed measurement module is linked with devices such as traffic lights and can automatically adjust signals when necessary to relieve traffic congestion. By integrating into the intelligent traffic management system, the speed measurement module helps improve the overall traffic efficiency and reduce traffic accidents and delays caused by slow driving. The implementation of this system not only optimizes the traffic flow, but also provides a strong basis for further traffic law enforcement, promoting the development of dynamic intelligent traffic management.

[0045] Further, the speed measurement of vehicles on the key sections of the fast lane includes the following steps:

[0046] S1. Data collection: Install a high-definition camera on the key section to monitor passing vehicles in real time; the camera covers the entire lane from the start detection line to the end detection line and has good night vision ability to ensure accurate capture of high-definition video streams under various lighting conditions; this step is the basis of the system, providing a reliable information source for subsequent analysis and monitoring through high-quality video data.

[0047] S2. Object detection: Use the YOLO algorithm to process the collected video stream to detect and identify target truck vehicles. This process involves using a pre-trained model to identify vehicles in each frame and track their movement trajectories. Through accurate object detection, the system can collect information about the driving conditions of each vehicle, providing the necessary data basis for speed calculation.

[0048] Further, the use of the YOLO algorithm to process the collected video stream to detect and identify vehicles includes the following steps:

[0049] S21. Data preparation: First, it is necessary to collect image or video data containing vehicles. Commonly used data sets include COCO and Pascal VOC. Then, use annotation tools to annotate the data, draw the bounding boxes of vehicles, and label each box with the corresponding label to generate the annotation files required for training.

[0050] S22. Data Preprocessing: In the data preprocessing stage, the labeled data is converted into the format required by YOLO, usually a text file where each line contains the class of a target and the relative coordinates of the bounding box. Additionally, data augmentation can be performed, such as random cropping, rotation, and brightness adjustment, to increase data diversity and thereby improve the robustness of the model.

[0051] S23. Environment Configuration: Next, the training environment needs to be configured by installing deep learning frameworks such as TensorFlow or PyTorch. Open-source implementations (such as YOLOv5) can be utilized, and a pre-trained model can be downloaded as the initial weights to accelerate the convergence speed during training.

[0052] S24. Model Training: In the model training stage, the training configuration items need to be set, including the network structure, input image size, learning rate, and batch size, etc. Then, run the training script and monitor the loss value and performance metrics to ensure that the model gradually converges and achieves the expected accuracy.

[0053] S25. Model Evaluation: After training is completed, the model is evaluated using the test set. By calculating metrics such as mAP (mean Average Precision), Precision, and Recall, the performance of the model is judged. At the same time, the detection results are visualized to evaluate the cases of missed detections and false detections.

[0054] S26. Model Deployment: The trained model is exported into a format suitable for inference, such as.pt or.onnx. This enables the model to be applied to real-time detection scenarios, such as processing video streams or image data, to achieve fast vehicle detection.

[0055] S27. Post-processing: In the post-processing stage, non-maximum suppression (NMS) is applied to remove bounding boxes with high overlap, and only the box with the highest confidence is retained. Finally, the detection results are visualized by drawing the vehicle boxes and class labels on the image for easy analysis and display of the detection results.

[0056] S3. Speed Calculation: After a vehicle is detected, the system analyzes the change in the vehicle's position in consecutive frames to calculate the current driving speed. Usually, the speed is achieved by measuring the displacement of the vehicle within a specific time. The key to this step lies in accuracy and real-time performance to ensure that vehicles with speeds below the preset threshold can be detected in a timely manner, thereby identifying the behavior of slow-moving vehicles occupying lanes.

[0057] S4. Data Recording and Analysis: The system automatically records relevant information, including data such as vehicle speed, time, and location. These records will be stored in the database for future query and analysis. By analyzing historical data, traffic management departments can identify traffic flow patterns and formulate reasonable management strategies to optimize traffic flow.

[0058] Furthermore, the key sections include main roads, traffic intersections, entrances and exits of expressways and highways, congestion hotspots, mountain roads and slope areas, as well as construction or obstacle areas.

[0059] In this embodiment, the main road is the road with the largest traffic flow in the city, usually connecting various important areas, and is prone to traffic congestion and accidents. Therefore, monitoring on the main road can effectively improve traffic efficiency.

[0060] A traffic intersection is a place where multiple roads meet, and vehicles turn and change lanes frequently, making accidents prone to occur. Monitoring these areas can help manage traffic lights and strengthen traffic control to ensure safety.

[0061] In the entrance and exit areas of expressways and highways, the behaviors of vehicles accelerating and decelerating are very crucial. Monitoring these areas helps prevent traffic accidents and improve the overall safety of the road.

[0062] Congestion hotspots: Some historical congestion points, such as near schools, commercial areas, and around large event venues, where the traffic flow increases sharply during specific time periods. Through monitoring, responses can be made in advance.

[0063] Mountain roads and slope areas: These sections with complex terrains may cause traffic problems due to factors such as weather and road conditions. Monitoring these areas helps detect and respond to sudden traffic jams or accidents in a timely manner.

[0064] Construction or obstacle areas: In areas with traffic construction or obstacles, the driving speed of vehicles may be affected. Monitoring these sections can provide real-time information for traffic managers so that timely measures can be taken.

[0065] By monitoring these key sections, the urban traffic flow can be better analyzed and managed, improving road safety and traffic efficiency.

[0066] The capture module captures the scene and obtains capture video images when the vehicle speed is lower than the preset speed threshold.

[0067] In this embodiment, the capture module uses another batch of high-definition cameras for capturing, and captures the scene only when the recognized vehicle speed is lower than a preset speed threshold. The triggering mechanism of the capture module ensures that the cameras do not work frequently under normal traffic flow conditions, thereby reducing unnecessary data redundancy and storage pressure. Only when the recognized vehicle speed is lower than the preset threshold, the system will activate the camera for shooting, which can effectively focus on the situations that may need attention and ensure the reasonable use of monitoring resources. At the same time, the application of high-definition cameras also improves the clarity of the images and the ability to capture details, making subsequent analysis and processing more reliable.

[0068] Furthermore, for the captured video images, high-position and low-position cameras are used for capturing simultaneously; among them, the high-position camera is used to identify the traffic flow and vehicle distribution on the road, and the low-position camera is used to identify vehicle details; the vehicle details include vehicle license plates and vehicle types.

[0069] The analysis module analyzes the on-site traffic conditions through the captured video images;

[0070] In this embodiment, analyzing the traffic conditions on the express lane through the captured video images mainly focuses on the assessment of vehicle flow, vehicle speed, and violations. First, the number of vehicles entering the express lane within a specific time period can be counted to determine the usage status and peak period of the lane. If it is found that the vehicles in the lane are dense and the speed is significantly lower than the normal driving speed, it may indicate a serious situation of lane occupation. Secondly, analyze the driving behaviors of vehicles in the video to identify vehicles that illegally occupy the express lane. For example, if non-express lane vehicles frequently enter the express lane or make improper behaviors such as stopping or decelerating, it will cause congestion in the overall traffic flow. In addition, traffic jams, braking phenomena, and too-close vehicle distances caused by lane occupation can also be observed, all of which can reflect the usage efficiency of the express lane. The comprehensive analysis of these data will help identify possible traffic problems on the express lane and provide improvement suggestions for traffic management departments, such as strengthening the monitoring of violations, increasing the publicity of express lane usage rules, formulating reasonable traffic flow optimization measures, and ensuring the smoothness and safety of the express lane. These measures will help improve the overall traffic efficiency and reduce congestion caused by lane occupation.

[0071] Furthermore, the analysis of the on-site traffic conditions through the captured video images includes the following steps:

[0072] T1. Identify the vehicle distribution information behind the target freight vehicle on the express road. The vehicle distribution information includes the number of vehicles, vehicle speeds, and corresponding vehicle behaviors; the vehicle behaviors include accelerating, decelerating, changing lanes, and overtaking;

[0073] In this embodiment, by analyzing the video image capture to obtain the vehicle information behind the target truck on the expressway, the real-time monitoring and evaluation of the traffic conditions in this area can be achieved. First, the system can identify and count the number of vehicles behind the target truck. This data can reflect the vehicle density following the target truck at a specific time and help identify the traffic flow conditions. Second, measuring the vehicle speed is crucial for evaluating the smoothness of the traffic flow. Through video analysis, the real-time speed data of each following vehicle can be obtained and compared with the speed of the target truck. This process can help determine the smoothness of the road and the risk of traffic congestion. In addition, analyzing vehicle behavior is an important part of comprehensively understanding traffic dynamics. The system can monitor behaviors such as acceleration, deceleration, lane change, and overtaking of each vehicle. For example, when a vehicle behind accelerates, it may indicate that the driver wants to overtake; while deceleration may mean there is an obstacle or an emergency ahead. At the same time, through the detection of lane change behavior, it can be determined whether the vehicle behind attempts to overtake the target truck and the success of its lane change.

[0074] T2. Predict the probability of current traffic congestion according to the vehicle distribution information.

[0075] Further, in step T2, the predicting the probability of current traffic congestion according to the vehicle distribution information includes the following steps:

[0076] T21. Collect the traffic condition data of trucks occupying the express lane in the historical period. The traffic condition data includes vehicle distribution information and traffic congestion information; the traffic congestion information is divided into congestion and non-congestion.

[0077] T22. Determine the Bayesian network structure with the vehicle distribution information as the parent node and the traffic congestion information as the child node.

[0078] In this embodiment, constructing a Bayesian network based on vehicle distribution information and traffic congestion information can effectively reveal the causal relationship between the two. In this network, "vehicle distribution information" is used as the parent node, including data such as the number, type, and speed of vehicles in a certain area; while "traffic congestion information" is used as the child node, indicating the current congestion status and degree. The network structure shows that vehicle distribution directly affects the occurrence of traffic congestion: when the number of vehicles in a certain area increases or the vehicle type changes, it may lead to a decrease in traffic smoothness and thus trigger traffic congestion. Therefore, the Bayesian network can be represented as "vehicle distribution information" pointing to "traffic congestion information", forming a causal chain.

[0079] In addition, after establishing the network, it is necessary to define the conditional probabilities for relevant nodes, which can quantify the likelihood of traffic congestion occurring under specific vehicle distributions. By learning using historical data, the network can be continuously updated and optimized, thereby improving its prediction ability. This Bayesian network not only provides valuable decision-making support for traffic management departments but also helps to identify potential congestion sources in a timely manner, enabling the adoption of corresponding intervention strategies, such as adjusting traffic signals, issuing traffic warnings, and guiding vehicle diversion. Ultimately, with the help of this network, traffic management can more scientifically and effectively address road congestion problems, enhancing traffic flow and safety.

[0080] T23. Use maximum likelihood estimation to determine the conditional probability distribution between the parent node and child node parameters, and complete the training of the Bayesian network model;

[0081] In the training of the Bayesian network model, the maximum likelihood estimation (MLE) method can effectively determine the conditional probability distribution between the parent node and child node. Specifically, relevant vehicle distribution information and traffic congestion information data need to be collected first. These data can come from traffic monitoring systems, historical records, or real-time sensors. By analyzing this data, we can establish a model relationship between the vehicle distribution state and traffic congestion. During the MLE process, we aim to find the parameter values that can maximize the observed data, that is, the probability distribution of traffic congestion occurring given the vehicle distribution. This can be achieved by calculating the frequencies of traffic congestion conditions under different vehicle numbers, vehicle types, time periods, etc., and then constructing a conditional probability table (CPT).

[0082] After completing the parameter estimation, the model will be able to reflect how changes in vehicle distribution affect the degree of traffic congestion, enabling this Bayesian network to make more accurate inferences and predictions. By continuously optimizing this model, traffic management departments can update the conditional probabilities based on real-time data, providing the urban traffic system with dynamic response capabilities. Ultimately, this Bayesian network will play an important role in understanding and predicting the mechanism of traffic congestion, helping to achieve more efficient traffic management and planning, and improving the operation efficiency and safety of the overall traffic system.

[0083] Furthermore, the formula for the conditional probability distribution is:

[0084]

[0085] In the formula, P(U) represents the joint conditional probability of a certain parameter U of the child node under all parent node parameters B i ; P a (B i ) is the probability of a certain parameter U of the child node occurring under a certain parameter B of the parent node i .

[0086] T24. Input the current vehicle distribution into the trained Bayesian network model to predict the probability of current traffic congestion.

[0087] In this embodiment, the Bayesian network model is a graphical model based on probabilistic inference, which is widely used to represent and reason about the conditional dependencies between variables. It consists of a directed acyclic graph (DAG) and a conditional probability table (CPT). Each node in the graph represents a random variable, and the directed edges represent the causal relationships between variables. This structure ensures the logicality of causal inference and avoids circular dependencies. The conditional probability table for each node describes the probability distribution of that node given its parent nodes (the variables that directly affect it), helping us understand how the variables change under different conditions. The Bayesian network is particularly good at handling uncertainty and can calculate the posterior probability of the variables of interest through inference algorithms, thus making effective predictions and decisions. In addition, the Bayesian network allows learning based on actual data. By analyzing the data, it can construct a model and adjust the relationships between variables to optimize the conditional probability distribution. This makes the model have good adaptability in a dynamic environment. Its application fields are very wide, covering medical diagnosis, risk assessment, fault detection, and traffic flow prediction, etc. Especially in traffic management, the Bayesian network can handle complex real-time data and provide a scientific basis for traffic congestion, accident risk, and flow optimization.

[0088] The decision-making module determines whether there is a traffic violation for the target truck vehicle according to the traffic situation.

[0089] Further, the determining whether there is a traffic violation for the target truck vehicle according to the traffic situation is specifically as follows: when the probability of current traffic congestion exceeds a preset first congestion threshold, it is determined whether there is a traffic violation for the target truck vehicle.

[0090] In this embodiment, if the calculated congestion probability exceeds this threshold, it means that the traffic condition is relatively serious and may lead to changes in traffic control measures. At this time, the behavior of the target truck needs to be further observed. Specifically, if the target truck still performs improper operations in the case of severe congestion at this time, then it can be determined that the vehicle has a traffic violation. This judgment not only depends on the calculation of probability but also needs to combine the real-time position of the vehicle and relevant monitoring information to ensure the accuracy of the judgment.

[0091] Further, when the decision-making module determines that there is a traffic violation for the target truck vehicle, it reminds the target truck vehicle not to occupy the fast lane through the traffic indicator light and retains the captured video image as the basis for the violation.

[0092] In this embodiment, when it is determined that the target truck has a traffic violation, a series of measures can be taken for management and punishment. First, use traffic lights to issue a warning to the truck, prompting it not to occupy the express lane. By setting a flashing red light or other obvious signals, the driver's attention can be effectively attracted, prompting them to follow traffic rules. This visual reminder method is particularly important on high-traffic roads and can effectively reduce traffic accidents caused by violations. Secondly, capturing video images as evidence of violations is another key link. Recording the road behavior of the target truck through surveillance cameras will provide important evidence for subsequent violation handling. These video materials can not only be used to verify the authenticity of violations but also serve as a legal basis to ensure fairness when necessary.

[0093] Furthermore, when the probability of the current traffic congestion exceeds a preset second congestion threshold, automatically report to the traffic management department and request handling of the traffic congestion.

[0094] It should be noted that the second congestion threshold in this embodiment is greater than the first congestion threshold. When the traffic congestion level exceeds the preset second congestion threshold, the system can automatically send an alarm request to the traffic management department for timely handling and alleviation of the congestion situation. The relatively high setting of the second congestion threshold means that the system will only actively issue an alarm when the traffic is extremely congested, thereby reducing false alarms caused by minor congestion. Specifically, the system can judge the congestion level by real-time monitoring of traffic flow and vehicle speed. Once the monitored traffic state reaches the second congestion threshold, the system will automatically generate an alarm message containing detailed data such as the current traffic conditions, specific locations, congestion periods, and possible affected ranges. This information will be quickly sent to the traffic management department for relevant personnel to intervene in a timely manner.

[0095] The above is only a preferred embodiment of the present invention and does not impose any form of limitation on the present invention. Although the present invention has been disclosed as above in a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to the above-disclosed technical content to form equivalent embodiments with equivalent changes, but as long as they do not depart from the technical content of the present invention, any brief modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A road condition monitoring system based on video image analysis, characterized in that: It includes a speed measurement module, a capture module, an analysis module, and a decision-making module. The speed measurement module, capture module, analysis module, and decision-making module are communicatively connected. Among them: The speed measurement module is used to measure the speed of target freight vehicles on key sections of the fast lane; The capture module captures the scene to obtain captured video images when the vehicle speed is lower than a preset speed threshold; The analysis module analyzes the traffic conditions at the scene through the captured video images; The decision-making module determines whether there are traffic violations for the target freight vehicle according to the traffic conditions; The analysis of the traffic conditions at the scene through the captured video images includes the following steps: T1. Identify the vehicle distribution information behind the target freight vehicle on the fast road. The vehicle distribution information includes the number of vehicles, vehicle speeds, and corresponding vehicle behaviors. The vehicle behaviors include accelerating, decelerating, changing lanes, and overtaking; T2. Predict the probability of current traffic congestion according to the vehicle distribution information; In step T2, the prediction of the probability of current traffic congestion according to the vehicle distribution information includes the following steps: T21. Collect traffic condition data of freight vehicles occupying the fast lane in historical periods. The traffic condition data includes vehicle distribution information and traffic congestion information. The traffic congestion information is divided into congested and uncongested; T22. Determine the Bayesian network structure with vehicle distribution information as the parent node and traffic congestion information as the child node; T23. Use maximum likelihood estimation to determine the conditional probability distribution between the parent node and child node parameters to complete the training of the Bayesian network model; T24. Input the current vehicle distribution into the trained Bayesian network model to predict the probability of current traffic congestion; The conditional probability distribution has the following calculation formula: ， In the formula, P(U) represents the joint conditional probability of a certain parameter U of the child node under all parent node parameters Bi; Pa(Bi) is the probability of a certain parameter U of the child node occurring under a certain parameter Bi of the parent node.

2. The road condition monitoring system based on video image analysis according to claim 1, wherein: The speed measurement of vehicles on key sections of the fast lane includes the following steps: S1. Data collection: Install high-definition cameras at key sections, and the cameras cover the entire lane from the start detection line to the end detection line; S2. Object detection: Use the YOLO algorithm to process the collected video stream to detect and identify target freight vehicles; S3. Speed calculation: After detecting the vehicle, analyze the vehicle position changes in consecutive frames to calculate the current driving speed.

3. A road condition monitoring system based on video image analysis according to claim 1, characterized in that: The key sections include main roads, traffic intersections, entrances and exits of expressways and highways, congestion hotspots, mountain roads and slope areas, as well as construction or obstacle areas.

4. The road condition monitoring system based on video image analysis according to claim 2, characterized in that: The use of the YOLO algorithm to process the collected video stream for vehicle detection and identification includes the following steps: S21. Data preparation: Collect image or video data containing vehicles, use annotation tools to annotate the data, draw the bounding boxes of the vehicles, and label each box with the corresponding label to generate the annotation files required for training; S22. Data preprocessing: Convert the annotated data into text file format; S23. Environment configuration: Configure the training environment and install the TensorFlow or PyTorch deep learning framework; S24. Model training: Set the configuration items for training, where the configuration items include network structure, input image size, learning rate, and batch size, and then run the training script, monitor the loss value and performance metrics to ensure that the model gradually converges and reaches the expected accuracy; S25. Model evaluation: After completing the training, use the test set to evaluate the model, and judge the performance of the model by calculating mAP, Precision, and Recall metrics.

5. The road condition monitoring system based on video image analysis according to claim 1, characterized in that: The captured video images are captured simultaneously by high and low cameras; among them, the high camera is used to identify the traffic flow and vehicle distribution on the road, and the low camera is used to identify vehicle details; the vehicle details include vehicle license plates and vehicle types.

6. The road condition monitoring system based on video image analysis according to claim 1, characterized in that: According to the traffic conditions, determine whether the target truck has traffic violations, specifically: when the probability of the current traffic congestion exceeds a preset first congestion threshold, determine whether the target truck has traffic violations.

7. A road condition monitoring system based on video image analysis according to claim 1, characterized in that: The decision-making module, when it is determined that the target truck has traffic violations, reminds the target truck not to occupy the fast lane through the traffic indicator light, and retains the captured video image as evidence of the violation.

8. The road condition monitoring system based on video image analysis according to claim 1, characterized in that: When the probability of the current traffic congestion exceeds a preset second congestion threshold, automatically report to the traffic management department and request to handle the traffic congestion.

Images