Vehicle behavior recognition gate opening system based on artificial intelligence

Abstract

The invention provides a vehicle behavior recognition gate opening system based on artificial intelligence, relates to the technical field of artificial intelligence, and aims to solve the problem that a traditional gate opening system is difficult to control gate opening through complex vehicle behaviors. The system comprises a data acquisition module, a vehicle behavior analysis module, a legal behavior judgment module and an intelligent module decision module. The data acquisition module integrates multi-source data from a camera, a laser radar and an infrared sensor, improves the image quality through a GAU-YOLOv5 model, and recognizes complex vehicle behaviors. The behavior legality judgment module performs judgment in combination with current vehicle behaviors, traffic rules and environmental conditions, and the intelligent decision module optimizes a gate opening strategy through a reinforcement learning algorithm to realize automatic control. According to the system, the vehicle behavior recognition precision and traffic management in a complex environment are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a vehicle behavior recognition and gate opening system based on artificial intelligence. Background Art

[0002] In the existing intelligent transportation system, gate opening control is an important link to ensure the orderly passage of vehicles; however, traditional gate opening systems usually rely on simple sensing equipment or manual control, which is difficult to cope with complex traffic conditions and changing environmental conditions; especially when faced with a large number of vehicles passing, these systems are often unable to accurately identify and judge the behavior of vehicles, resulting in false gates, wrong gates, etc., which not only affects traffic efficiency, but also poses certain safety hazards.

[0003] With the development of artificial intelligence technology, more and more gate opening systems have begun to introduce intelligent decision-making mechanisms, such as using machine learning algorithms to analyze and judge vehicle behavior; however, in the existing technology, these systems still have deficiencies in dealing with the identification and legality judgment of complex vehicle behaviors in complex weather conditions; traditional vehicle behavior recognition algorithms can usually only process a single type of sensor data and have difficulty fusing multi-source data, resulting in low recognition accuracy, especially in complex weather or complex road conditions; most of the existing gate opening systems lack real-time learning and adjustment mechanisms and are unable to dynamically adjust the gate opening according to the changing environment. Summary of the invention

[0004] 1. Technical issues to be solved

[0005] In view of the deficiencies in the prior art, the present invention provides a vehicle behavior recognition and gate opening system based on artificial intelligence, which solves the problem of how to accurately identify vehicle behavior in complex environments and control the gate opening by judging the vehicle behavior.

[0006] (II) Technical solution

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a vehicle behavior recognition and gate opening system based on artificial intelligence, comprising:

[0008] The data acquisition module is used to integrate the data collected by the monitoring equipment and transmit the data to the vehicle behavior analysis module through data integration;

[0009] The vehicle behavior analysis module uses the GAU-YOLOv5 model to improve the quality of vehicle images and identify complex vehicle behaviors based on the data integrated by the data acquisition module;

[0010] A behavior legality judgment module, used to judge the legality and rationality of the vehicle behavior identified by the vehicle behavior analysis module in combination with current traffic rules and environmental conditions;

[0011] The intelligent decision-making module optimizes the gate control strategy based on the results of the behavior legality judgment module through the reinforcement learning algorithm, and automatically determines whether to open the gate or delay opening the gate.

[0012] Preferably, the data acquisition module collects the behavior data of the vehicle through monitoring equipment, wherein the monitoring equipment includes a camera, an infrared sensor and a lidar; wherein the camera provides high-resolution visual data, the lidar captures three-dimensional point cloud data of the environment, and the infrared sensor collects thermal imaging data under low light or bad weather conditions; the module synchronizes the data of all monitoring devices in time, combines the data collected at the same time, and eliminates the time deviation; the data acquisition module denoises and preprocesses various types of data; for the lidar data, removes stray points; for the camera visual data, eliminates blur or light interference noise; the module adjusts the resolution of the data according to demand, and standardizes the image and three-dimensional point cloud data to a consistent resolution, and the data acquisition module converts the raw data of different sensors into a unified processing format, such as converting video data into frame images and point cloud data into two-dimensional or three-dimensional matrices for subsequent processing; after completing the preprocessing, the data acquisition module uses data fusion technology to integrate and process the data from different monitoring devices to form a unified data set; through feature-level fusion, the thermal characteristics of the infrared image are combined with the distance information of the lidar; the data acquisition module transmits the fused data to the vehicle behavior analysis module.

[0013] Preferably, the data fusion technology uses key features in the data set, including visual features of the camera, three-dimensional spatial features of the lidar, and thermal features of the infrared sensor; the key features are merged together through a feature fusion algorithm, and these features are converted into a unified format and scale; in the feature fusion stage, the system adopts a feature splicing method to assign weights according to the importance of data features for judging vehicle behavior, so as to achieve effective integration of different features; the standardized features are spliced ​​according to the weights to create a unified feature, in which the contribution of each feature is proportional to its weight; for example, if there are two features A and B, with weights wA and wB respectively, then the spliced ​​features can be expressed as: FeatureVector = [wA*A, wB*B]; the fused features are processed by the GAU-YOLOv5 model.

[0014] Preferably, the vehicle behavior analysis module improves image quality and identifies vehicle behavior through GAU-YOLOv5; integrates the intermediate feature layer of the GAN generator into the feature extraction network of YOLOv5, provides YOLOv5 with richer context information and high-quality feature maps, and enhances its understanding and recognition capabilities of vehicle images; through the feature cascade fusion technology, the corresponding layer of the GAN generator is cross-layered with each layer of the convolutional network of YOLOv5, further improving YOLOv5's understanding of local and global features of the image, especially for vehicle behavior detection under complex weather conditions; the GAN discriminator evaluates the quality of the generated image and provides feedback information to the detection module of YOLOv5 to help it identify key features in target detection and Prioritize processing. During the training process, through joint loss function optimization, the discriminator of GAN and YOLOv5 share the same loss function, which integrates the generation loss, discrimination loss and detection loss of YOLOv5, and realizes the simultaneous improvement of GAN generation quality and YOLOv5 detection performance through back propagation; using the high-quality feature map provided by the GAN generator, YOLOv5 uses high-quality feature maps to identify clearer features to judge the dynamic behavior of the vehicle, including sudden braking, reverse driving, and lane changing; combined with the image authenticity judgment provided by the discriminator, YOLOv5 further improves the judgment accuracy of vehicle behavior; after identifying the complex behavior of the vehicle, the rationality and legality of the vehicle behavior are jointly evaluated through the linkage between YOLOv5 and the discriminator of GAN

[0015] Preferably, the behavior legality judgment module receives the output from the vehicle behavior analysis module, and the behavior legality judgment module matches the vehicle behavior with the traffic rules; if the vehicle's driving speed exceeds the road standard, the module marks the behavior as illegal; analyzes the environmental conditions, which include the current weather conditions, road conditions, density of surrounding vehicles and driving conditions; environmental data is obtained through monitoring equipment and input into the legality judgment module; the analysis of environmental conditions enables the system to better understand whether certain behaviors are reasonable in a complex environment; the behavior legality judgment module verifies each piece of behavior data step by step through logical reasoning and conditional judgment; for each type of vehicle behavior, the behavior legality judgment module generates a legality and rationality score, and generates a decision result of the behavior legality judgment module based on the degree of match between the score and the traffic rules and environmental conditions, and the decision result includes the legality mark and rationality evaluation report of the specific behavior, which will be used as input to the subsequent intelligent decision module for further action decisions.

[0016] Preferably, the intelligent decision-making module receives the decision results from the vehicle behavior analysis module, and the intelligent decision-making module converts the decision results into the observation state of the reinforcement learning algorithm, and the observation state reflects the comprehensive situation of the current traffic environment and vehicle behavior; the intelligent agent in the intelligent decision-making module selects an action according to the observation state and the predefined strategy, wherein the action is a decision to open the gate or delay the gate opening; the intelligent agent performs the corresponding gate control operation according to the selected action, and observes the resulting results, such as traffic flow, vehicle waiting time, etc.; the observation results are converted into reward signals and fed back to the intelligent agent, and the reward signal is designed to encourage the intelligent agent to choose actions that can improve traffic efficiency and ensure safety; the intelligent agent uses these reward signals to update its strategy, and continuously optimizes the strategy through the reinforcement learning algorithm to maximize the long-term cumulative reward; as time goes by, the intelligent decision-making module makes the most appropriate gate control decision in different situations through continuous learning and strategy updating; the intelligent decision-making module makes a quick decision on whether to open the gate by analyzing vehicle behavior and environmental conditions in real time.

[0017] (III) Beneficial effects

[0018] The present invention provides a vehicle behavior recognition and gate opening system based on artificial intelligence, which has the following beneficial effects:

[0019] 1. The present invention can effectively improve the vehicle image quality and behavior recognition accuracy in complex environments by adopting the GAU-YOLOv5 model, especially in severe weather and complex traffic conditions, and significantly reduce recognition errors.

[0020] 2. The data acquisition module of the present invention can integrate multi-source data from cameras, lidars and infrared sensors, and through data fusion, comprehensively obtain various behavioral characteristics of the vehicle, thereby enhancing the system's comprehensive perception of vehicle behavior.

[0021] 3. The present invention analyzes the behavior of vehicles in complex weather conditions, determines the rationality and legality of the vehicle's behavior, and enables the system to control gate opening based on the rationality and legality of the vehicle's behavior, thereby improving the overall intelligent management level of traffic, enhancing traffic safety, optimizing traffic smoothness, and improving the effectiveness of traffic control under complex weather conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Embodiment 1:

[0025] The embodiment of the present invention provides a vehicle behavior recognition and gate opening system based on artificial intelligence, including a data acquisition module, a vehicle behavior analysis module, a behavior legality judgment module and an intelligent decision-making module.

[0026] The data acquisition module collects the behavior data of the vehicle through monitoring equipment, which includes cameras, infrared sensors and laser radars. The cameras provide high-resolution visual data, the laser radars capture the three-dimensional point cloud data of the environment, and the infrared sensors collect thermal imaging data under low light or bad weather conditions. The module synchronizes the data of all monitoring devices in time, combines the data collected at the same time, and eliminates the time deviation. The data acquisition module denoises and preprocesses various types of data. For laser radar data, it removes stray points; for camera images, it eliminates blur or light interference noise. The module adjusts the resolution of the data according to demand and standardizes the image and three-dimensional point cloud data to a consistent resolution. The data acquisition module converts the raw data of different monitoring devices into a unified processing format, such as converting video data into frame images and point cloud data into two-dimensional or three-dimensional matrices for subsequent processing. After completing the preprocessing, the data acquisition module uses data fusion to integrate and process the data from different monitoring devices to form a unified data set. Through feature-level fusion, the thermal characteristics of the infrared image are combined with the distance information of the laser radar. The data acquisition module transmits the fused data to the vehicle behavior analysis module through the data transmission protocol.

[0027] Data fusion technology extracts key features from the integrated data set, including visual features of the camera, three-dimensional spatial features of the lidar, and thermal features of the infrared sensor; the key features are merged together through a feature fusion algorithm and converted into a unified format and scale; in the feature fusion stage, the system uses feature splicing to assign weights based on the importance of data features for judging vehicle behavior to achieve effective integration of different features; the standardized features are spliced ​​according to the weights to create a unified feature, in which the contribution of each feature is proportional to its weight; for example, if there are two features A and B, with weights wA and wB respectively, the spliced ​​features can be expressed as: FeatureVector = [wA*A, wB*B]; the fused features are processed by the GAU-YOLOv5 model.

[0028] The vehicle behavior analysis module uses GAU-YOLOv5 to improve image quality and identify vehicle behavior; the intermediate feature layer of the GAN generator is integrated into the feature extraction network of YOLOv5 to provide YOLOv5 with richer contextual information and high-quality feature maps, enhancing its understanding and recognition capabilities of vehicle images; through the feature cascade fusion technology, the corresponding layer of the GAN generator is cross-connected with each layer of the convolutional network of YOLOv5, further improving YOLOv5's understanding of local and global information of the image, especially the vehicle behavior under complex weather conditions. For detection; GAN's discriminator evaluates the quality of the generated image and provides feedback information to YOLOv5's detection module to help it identify key features in target detection and perform priority processing. During the training process, through joint loss function optimization, GAN's discriminator and YOLOv5 share the same loss function, which integrates generation loss, discrimination loss and YOLOv5's detection loss, and realizes simultaneous improvement of GAN generation quality and YOLOv5 detection performance through back propagation; the loss function includes generation loss, discrimination loss and detection loss. The generation loss is used to evaluate the difference between the image generated by the GAN generator and the real image, so that the quality of the generated image is close to the real image; the discrimination loss promotes the improvement of the GAN discriminator, so that it can accurately evaluate the quality of the generated image, thereby providing more accurate feedback for YOLOv5; the detection loss directly optimizes the performance of YOLOv5 to ensure its effect on target detection tasks; the mathematical expression of the loss function is Ltotal=λ1Lgen+λ2Ldis+λ3Ldet, where Lgen is the generation loss, calculated as the difference between the generated image and the real image is a measure of difference; Ldis is the discriminative loss, which consists of two parts: one part evaluates the discriminator's ability to recognize real images, and the other part evaluates the discriminator's ability to recognize generated images; Ldet is the detection loss, including classification loss, positioning loss and confidence loss, which respectively evaluate the difference between the predicted category, bounding box and confidence and the actual situation; λ1, λ2, λ3 are corresponding weight coefficients used to balance the contribution of different losses; through the back-propagation algorithm, the gradient of the loss function is calculated and used to update the parameters of the GAN generator and discriminator, as well as the network weights of YOLOv5.

[0029] Using the high-quality feature map provided by the GAN generator, YOLOv5 uses the high-quality feature map to identify clearer features to judge the dynamic behavior of the vehicle. YOLOv5 extracts features from the high-quality feature map and extracts the key features of the vehicle, such as the shape, size, color, and texture of the vehicle through its internal convolutional layer and pooling layer. With the completion of feature extraction, YOLOv5 further uses the key features to identify the vehicle's behavior. At this stage, the network structure of YOLOv5 identifies the dynamic behavior of the vehicle, including sudden braking, reverse driving, lane changing, etc. The identification of vehicle behavior is based on the vehicle's motion pattern in the image and its relative position relationship with other vehicles or objects. For example, if a vehicle is in If there is a sudden deceleration or change of direction in the image, YOLOv5 will determine the corresponding behavior based on the key features; then, the GAN discriminator intervenes to evaluate the authenticity of the vehicle behavior recognized by YOLOv5; the discriminator is trained to distinguish between real images and generated images, and its output reflects the credibility of YOLOv5's recognition results; if the discriminator believes that YOLOv5's recognition results are reliable, then this behavior recognition will be further confirmed; if the discriminator believes that the result is questionable, the system will re-evaluate and require more information or more in-depth analysis to determine the authenticity of the behavior; after confirming the authenticity of the vehicle's behavior, the combination of YOLOv5 and GAN discriminators accurately identifies the complex behavior of the vehicle.

[0030] The behavior legality judgment module receives the vehicle behavior output from the vehicle behavior analysis module, and matches the vehicle behavior output with the traffic rules. If the vehicle's driving speed exceeds the road standard, the behavior legality judgment module marks the behavior as illegal; analyzes the environmental conditions, which include the current weather conditions, road conditions, density of surrounding vehicles and driving conditions; environmental data is obtained through monitoring equipment and input into the legality judgment module; the analysis of environmental conditions enables the system to better understand whether certain behaviors are reasonable in complex environments; consider the rationality of vehicle behavior, which may be unreasonable under current environmental conditions even if it is legal; the behavior legality judgment module verifies each piece of behavior data step by step through logical reasoning and conditional judgment; for each type of vehicle behavior, the behavior legality judgment module generates a legality and rationality score, and based on the degree of matching between the score and traffic rules and environmental conditions, generates the decision result of the behavior legality judgment module, including the legality mark and rationality evaluation report of the specific behavior, and the decision result will be used as the input of the intelligent decision module for further gate opening action control.

[0031] The intelligent decision-making module receives data from the vehicle behavior analysis module and the legitimacy and rationality evaluation results provided by the behavior legitimacy judgment module; the intelligent decision-making module converts these data into the observation state of the reinforcement learning algorithm, and the observation state reflects the comprehensive situation of the current traffic environment and vehicle behavior; the intelligent agent in the intelligent decision-making module selects an action based on the observation state and the predefined strategy, where the action is the decision to open the gate or delay the gate opening; the intelligent agent performs the corresponding gate control operation according to the selected action and observes the resulting results, such as traffic flow, vehicle waiting time, etc. These observation results are converted into reward signals and fed back to the intelligent agent. The intelligent agent continuously optimizes its strategy through the reinforcement learning algorithm. As time goes by, the intelligent decision-making module gradually learns to make the most appropriate gate control decisions in different situations through continuous learning and strategy updating. For example, if the vehicle behavior is legal and the traffic environment permits, the intelligent agent chooses to open the gate; if the vehicle is driving fast in foggy weather, the intelligent agent chooses to delay opening the gate to wait for a better time or ensure safety. The intelligent decision-making module can realize highly automated gate control. By analyzing vehicle behavior and traffic conditions in real time, it can quickly make decisions on whether to open the gate, so as to optimize traffic flow and reduce human errors, thereby improving the intelligence level and efficiency of the traffic management system.

[0032] Embodiment 2:

[0033] The vehicle behavior recognition and gate opening system based on artificial intelligence of the present invention is applied in a complex urban traffic environment.

[0034] First, cameras, lidars, and infrared sensors are deployed at intersections to collect multi-source data on vehicles in real time. Cameras acquire high-resolution visual images, lidars generate three-dimensional point cloud data of vehicles and their surroundings, and infrared sensors collect thermal imaging data in low-light or bad weather conditions. Time synchronization is first performed to ensure that data from different monitoring devices can be associated at the same time point. Then, the data acquisition module synchronizes various types of data, removes stray points in lidar data, eliminates noise or light interference in images, and standardizes the resolution of the data according to the requirements of the GAU-YOLOv5 model.

[0035] After preprocessing, the system integrates the data from different monitoring devices into a unified data set through data fusion technology. The data set integrates the visual features of the camera, the spatial features of the lidar, and the thermal features of the infrared sensor; the fused data set is transmitted to the vehicle behavior analysis module. After receiving this data, the GAU-YOLOv5 model first improves the image quality and then identifies the specific behavior of the vehicle. The GAN generator is combined with the feature extraction layer of the YOLOv5 model to enhance the system's recognition ability in complex environments; the model identifies complex driving behaviors such as sudden braking, reversing or irregular lane changes by analyzing the vehicle's motion trajectory, relationship with road markings, vehicle speed, and headlight status.

[0036] The identified vehicle behaviors are then passed to the behavior legality judgment module, which compares these behaviors with current traffic regulations and real-time environmental conditions. If the vehicle's behavior, such as speeding or illegal lane changes, conflicts with traffic regulations, the system will mark the behavior as illegal. In addition, the behavior legality judgment module analyzes the current road conditions, weather conditions, and the density of surrounding traffic to determine whether certain vehicle behaviors are reasonable in the current environment, even if they are legal. Finally, the behavior legality judgment module generates a legality and rationality score, and passes these scores together with the behavior recognition results to the intelligent decision-making module.

[0037] Based on the above data, the intelligent decision-making module uses a reinforcement learning algorithm to generate a gate opening strategy; the agent in this module chooses whether to open the gate or delay opening the gate according to the legality and rationality scores of the vehicle behavior and the comprehensive situation of the current traffic environment; if the vehicle behavior complies with traffic regulations and environmental conditions permit, the system will open the gate immediately; if there is a risk in the vehicle behavior, such as too many vehicles and a small gate in foggy weather, the system will choose to delay opening the gate to ensure traffic safety; throughout the process, the agent continuously learns from the operation results, optimizes its decision-making strategy through feedback, and gradually improves the system's traffic management efficiency and safety; ultimately, the intelligent decision-making module can achieve real-time gate opening control in complex traffic environments to ensure safe and smooth vehicle traffic.

[0038] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A vehicle behavior recognition and gate opening system based on artificial intelligence, characterized in that: include: The data acquisition module is used to integrate the data collected by the monitoring equipment and transmit the data to the vehicle behavior analysis module through data integration; The vehicle behavior analysis module uses the GAU-YOLOv5 model to improve the quality of vehicle images and identify complex vehicle behaviors based on the data integrated by the data acquisition module; A behavior legality judgment module, used to judge the legality and rationality of the vehicle behavior identified by the vehicle behavior analysis module in combination with current traffic rules and environmental conditions; The intelligent decision-making module optimizes the gate control strategy based on the results of the behavior legality judgment module through the reinforcement learning algorithm, and automatically determines whether to open the gate or delay opening the gate.

2. The vehicle behavior recognition and gate opening system based on artificial intelligence according to claim 1 is characterized by: The data acquisition module collects the behavior data of the vehicle through monitoring equipment, wherein the monitoring equipment includes a camera, an infrared sensor and a laser radar; The camera provides high-resolution visual data, the lidar captures the three-dimensional point cloud data of the environment, and the infrared sensor collects thermal imaging data in low light or bad weather conditions; the data acquisition module synchronizes the data of all monitoring devices, combines the data collected at the same time, and eliminates the time deviation; the data acquisition module pre-processes various types of data; for the lidar data, removes stray points; For camera visual data, remove blur or light interference noise; The data acquisition module adjusts the data resolution according to the requirements of the GAU-YOLOv5 model and standardizes the resolution of images and 3D point cloud data. The data acquisition module converts the raw data of different monitoring devices into a unified processing format. After completing the preprocessing, the data acquisition module uses data fusion to integrate the data from different monitoring devices to form a data set; The data acquisition module transmits the fused data set to the vehicle behavior analysis module.

3. The vehicle behavior recognition and gate opening system based on artificial intelligence according to claim 2 is characterized in that: The data fusion is performed by extracting key features of the data set, wherein the key features include visual features in the visual data of the camera, three-dimensional spatial features of the three-dimensional point cloud data in the laser radar, and thermal features of the thermal imaging data in the infrared sensor; The key features are merged together through feature fusion algorithms, and the key features of the data are converted into a unified format and scale; During the feature fusion stage, the system uses a feature splicing method to assign weights based on the importance of data features used to determine vehicle behavior, thereby achieving effective integration of different features.

4. The vehicle behavior recognition and gate opening system based on artificial intelligence according to claim 3 is characterized by: The vehicle behavior analysis module inputs the fused features into the GAU-YOLOv5 model for processing, improves the image quality and identifies the vehicle behavior through GAU-YOLOv5; integrates the intermediate feature layer of the GAN generator into the feature extraction network of YOLOv5, provides YOLOv5 with richer context information and high-quality feature maps, and enhances its understanding and recognition capabilities of vehicle images; cross-layer connections are made between the convolution layer of the GAN generator and each layer of the convolution network of YOLOv5, further improving YOLOv5's understanding of local and global features of the image, especially vehicle behavior detection under complex weather conditions; the GAN discriminator evaluates the quality of the generated image and provides feedback information to the detection module of YOLOv5, helping YOLOv5 detect and identify key features and perform priority processing.

5. The vehicle behavior recognition and gate opening system based on artificial intelligence according to claim 4 is characterized by: Using the high-quality feature map provided by the GAN generator, YOLOv5 enters the vehicle behavior recognition stage. In this stage, YOLOv5 processes the extracted high-quality feature map through its detection head, generates a prediction box and a confidence score for identifying and locating the vehicle behavior in the image, and outputs the vehicle behavior. The prediction box is used to indicate the position and range of the vehicle, and the confidence score is between 0 and 1, indicating the degree of certainty of the GAU-YOLOv5 model that the prediction box contains the vehicle behavior. During the prediction process, YOLOv5 judges the vehicle behavior by analyzing the local and global features in the high-quality feature map. After identifying the potential complex vehicle behavior, YOLOv5 further combines the feedback of the GAN discriminator to improve the accuracy of the behavior. The GAN discriminator evaluates the authenticity and consistency of the generated feature map and the recognition result. Through the GAN discriminator, YOLOv5 eliminates misjudgments caused by image noise or other interference.

6. The vehicle behavior recognition and gate opening system based on artificial intelligence according to claim 5 is characterized by: The behavior legality judgment module receives the output from the vehicle behavior analysis module, and the behavior legality judgment module matches the vehicle's behavior with traffic rules; if the vehicle's driving speed exceeds the road standard, the behavior legality judgment module marks the behavior as illegal; analyzes environmental conditions, which include current weather conditions, road conditions, density of surrounding vehicles and driving conditions; environmental data is obtained through monitoring equipment and input into the legality judgment module; the behavior legality judgment module verifies the behavior of each vehicle step by step through logical reasoning and conditional judgment; for each behavior of the vehicle, the behavior legality judgment module generates a legality and rationality score, and generates a decision result of the behavior legality judgment module based on the degree of matching between the score and traffic rules and environmental conditions, and the decision result includes the legality mark and rationality evaluation report of the specific behavior.

7. The vehicle behavior recognition and gate opening system based on artificial intelligence according to claim 6 is characterized by: The intelligent decision-making module receives the decision results from the vehicle behavior analysis module and converts them into the observation state of the reinforcement learning algorithm, and the observation state reflects the comprehensive situation of the current traffic environment and vehicle behavior; the intelligent agent in the intelligent decision-making module selects an action according to the observation state and the predefined gate opening and closing strategy, wherein the action includes the decision of opening, closing or delaying the gate opening; the intelligent agent performs the corresponding gate control operation according to the selected action and observes the resulting results; these observation results are converted into reward signals and fed back to the intelligent agent; the intelligent agent selects the most appropriate gate control decision by accumulating reward signals; the intelligent decision-making module analyzes vehicle behavior and traffic conditions in real time through automated gate control, and quickly makes a decision on whether to open the gate.

Images