Intelligent decision support method and system for vehicle-mounted sensor data fusion

Abstract

The present invention relates to the field of vehicle data processing, and in particular to an intelligent decision support method and system for vehicle-mounted sensor data fusion. The method comprises the following steps: obtaining vehicle multi-source operating status monitoring data based on vehicle-mounted sensors; performing vehicle status heterogeneous fusion and operating status multi-period change analysis on the vehicle multi-source operating status monitoring data, thereby obtaining vehicle operating status evolution trend characteristics; obtaining vehicle driving recorder monitoring images and vehicle dynamic environment monitoring parameters, and performing multi-target visual detection and dynamic operating scene fitting to construct a vehicle operating status scene graph; performing three-dimensional scene point cloud modeling on the vehicle operating status scene graph, and performing dynamic vehicle trend rendering according to the vehicle operating status evolution trend characteristics, thereby obtaining a dynamic vehicle scene twin model. The present invention realizes efficient and intelligent vehicle data decision analysis, and improves the safety and stability of vehicle driving.

Description

Technical Field

[0001] The present invention relates to the field of vehicle data processing, and in particular to an intelligent decision support method and system for vehicle-mounted sensor data fusion. Background Art

[0002] With the rapid development of intelligent transportation systems (ITS) and autonomous driving technologies, the application of vehicle sensors has become an important technical foundation for improving vehicle safety, comfort and automation. Modern cars are equipped with a variety of sensors, such as radar, lidar, camera, inertial measurement unit (IMU), GPS, ultrasonic sensor, etc., which are used to monitor the vehicle's surrounding environment, driving status and road conditions in real time. These sensors can provide multi-dimensional information about vehicle position, speed, acceleration, obstacle distance, lane departure, etc., and become a key data source for realizing intelligent driving, assisted driving and vehicle health management.

[0003] However, data from a single sensor often suffers from problems such as insufficient accuracy, missing data, and noise interference in complex and changing traffic environments. For example, although radar can provide relatively stable distance measurements under adverse weather conditions, it has certain limitations in detection accuracy and object recognition; cameras can provide high-resolution images, but their recognition capabilities are weak in low light and adverse weather conditions; and inertial sensors can provide accurate vehicle dynamic information, but their perception of the external environment is relatively weak. Therefore, traditional vehicle status monitoring and decision analysis methods often rely on a single data source, making it difficult to guarantee the accuracy and reliability of the analysis results. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes an intelligent decision support method and system for vehicle-mounted sensor data fusion to solve at least one of the above technical problems.

[0005] To achieve the above object, the present invention provides an intelligent decision support method for vehicle-mounted sensor data fusion, comprising the following steps: Step S1: acquiring vehicle multi-source operating status monitoring data based on vehicle-mounted sensors; performing vehicle state heterogeneous fusion and operating state multi-period change analysis on the vehicle multi-source operating status monitoring data, thereby obtaining vehicle operating state evolution trend characteristics; Step S2: Acquire the vehicle driving recorder monitoring image and vehicle dynamic environment monitoring parameters, perform multi-target visual detection and dynamic operation scene fitting, and construct a vehicle operation scene diagram; Step S3: Modeling the three-dimensional scene point cloud of the vehicle running state scene graph, and rendering the dynamic vehicle trend according to the evolution trend characteristics of the vehicle running state, so as to obtain a dynamic vehicle scene twin model; Step S4: Perform dynamic vehicle operation simulation on the dynamic vehicle scenario twin model, and then perform dynamic emergency risk analysis to generate scenario dynamic emergency risk data; Step S5: Perform dynamic emergency avoidance behavior decision and safe driving path planning on the scene dynamic emergency risk data, thereby generating a visual view of the emergency risk decision; Step S6: Based on the emergency risk decision visualization view, instant perception feedback is performed, and vehicle global self-optimization is performed to build a vehicle intelligent global decision engine.

[0006] The present invention obtains more comprehensive and accurate vehicle operation status information by integrating data from different vehicle-mounted sensors (such as radar, camera, GPS, IMU, etc.). The data from traditional single sensors is difficult to fully reflect the actual situation of the vehicle, and multi-source data fusion can effectively make up for their respective shortcomings. The vehicle status information is usually collected by different sensors, and the dimensions, forms and timeliness of this information are different. Through heterogeneous fusion technology, data from different sources can be integrated into a unified and comprehensive vehicle operation status model to improve the accuracy and robustness of state estimation. Using the images and parameters collected by the vehicle driving recorder and environmental monitoring sensors, it can detect and identify various surrounding targets (such as other vehicles, pedestrians, obstacles, etc.) in real time. This provides basic data for the vehicle's intelligent perception and enhances the vehicle's perception ability in complex traffic environments. Based on the visual detection results, a real-time vehicle operating environment scene is constructed through a fitting algorithm, which provides an intuitive and accurate environmental model for subsequent decision-making, and can provide real-time response and decision-making support for vehicles in dynamically changing traffic scenes. By constructing a three-dimensional scene point cloud model, the system can present the spatial structure of the vehicle's surrounding environment and provide richer spatial information than traditional two-dimensional images. The point cloud model effectively describes the depth information of the environment and provides accurate spatial positions for obstacles and driving trajectories in complex scenes. Combined with the evolution trend of the vehicle's operating state, the vehicle's future driving trajectory is predicted and dynamically rendered, which helps to identify potential obstacles in advance. It optimizes driving paths and supports simulation testing and dynamic evaluation. Through simulation technology, it simulates the vehicle's motion state and its interaction with the surrounding environment in different driving scenarios in a virtual environment. This simulation can detect potential problems in advance, evaluate the vehicle's performance in different scenarios, and provide data support for safety evaluation. Through emergency risk analysis of dynamic vehicle operation data, the system can identify the risks faced by vehicles in complex scenarios, such as emergency braking, obstacle avoidance, collision warning, etc. Based on risk analysis, it can generate scene dynamic emergency risk data to warn of potential accidents in advance. According to the dynamic emergency risk data, the system can intelligently generate emergency avoidance behaviors, such as choosing to avoid obstacles, adjust driving speed, change lanes, etc. This can It can reduce the driver's stress in emergencies and reduce the probability of accidents. In emergencies, the system plans the safest driving route based on the current scenario and risk analysis results. This route planning not only takes into account the current state of the vehicle, but also adapts to changes in the traffic environment in real time. Through instant perception feedback on the visual decision view, the system can quickly adjust the driving strategy. This feedback mechanism helps to improve the vehicle's response speed and decision-making efficiency, and enhance the system's adaptability to emergencies. The vehicle's intelligent global decision-making engine can integrate various perception data, operating status and environmental information to perform global strategy optimization. This optimization can continuously adjust the vehicle's behavior to ensure that it always maintains the best decision-making level in different driving environments.

[0007] In this specification, an intelligent decision support system for vehicle-mounted sensor data fusion is provided, which is used to execute the intelligent decision support method for vehicle-mounted sensor data fusion as described above, including: The heterogeneous fusion module is used to obtain multi-source vehicle operation status monitoring data based on vehicle-mounted sensors; the vehicle state heterogeneous fusion and multi-period change analysis of the vehicle operation status monitoring data are performed to obtain the evolution trend characteristics of the vehicle operation status; The running scene module is used to obtain the vehicle driving recorder monitoring image and vehicle dynamic environment monitoring parameters, and perform multi-target visual detection and dynamic running scene fitting to construct the vehicle running scene diagram; The twin model module is used to perform three-dimensional scene point cloud modeling on the vehicle operation scene graph, and perform dynamic vehicle trend rendering according to the evolution trend characteristics of the vehicle operation state, thereby obtaining a dynamic vehicle scene twin model; The operation simulation module is used to perform dynamic vehicle operation simulation on the dynamic vehicle scenario twin model, and then perform dynamic emergency risk analysis to generate scenario dynamic emergency risk data; The emergency decision-making module is used to make dynamic emergency avoidance behavior decisions and safe driving path planning based on the scene dynamic emergency risk data, thereby generating a visual view of the emergency risk decision; The instant feedback optimization module is used to provide instant perception feedback based on the emergency risk decision visualization view, perform global self-optimization of the vehicle, and build an intelligent global decision-making engine for the vehicle.

[0008] The present invention can obtain data of different dimensions from different types of vehicle-mounted sensors (such as radar, camera, GPS, acceleration sensor, etc.) and fuse them through the heterogeneous fusion module. This means that the system can take advantage of the data of various sensors to form a full range of more comprehensive vehicle status information. By analyzing the operation data of multiple time periods, the module identifies the change trend of vehicle status over time (such as speed change, driving habits, driving environment, etc.). The extraction of this trend feature helps to predict future vehicle status changes and provide data support for subsequent decision-making. Through the combination of driving recorders and dynamic environmental monitoring parameters, the system can perceive various factors on the road in real time (such as pedestrians, etc.). , other vehicles, road obstacles, etc.), so as to fully grasp the dynamic environment around the vehicle. Through multi-target visual detection, the module can identify multiple potential dangerous targets in dynamic scenes (such as other vehicles, pedestrians, traffic signs, etc.), and combine this information with the current driving scene. Such multi-target perception capability greatly enhances the real-time response capability of the system. The 3D scene point cloud modeling more realistically restores the environment in which the vehicle is located, including elements such as roads, surrounding vehicles, obstacles, etc. Through high-precision 3D modeling, the vehicle better understands and predicts changes in the surrounding environment. According to the evolution trend characteristics of the vehicle's operating state, the twin model module can render dynamic trends in real time and simulate different driving states. Through this rendering, the system predicts the driving path and potential dangers in advance and provides decision support. By running simulations, the system can simulate different driving scenarios, conduct virtual experiments, and identify potential traffic risks in advance. This process helps to evaluate the risks of different decision-making plans and ensure the selection of the best decision. During the simulation process, the system can identify and analyze the probability of various emergency situations (such as sudden braking of vehicles, collisions, etc.). Through emergency risk analysis, the system formulates appropriate emergency response strategies to improve driving safety. Based on real-time risk data, the system can make rapid emergency decisions (for example: automatic braking, obstacle avoidance, lane changes, etc.) to ensure that the vehicle can respond quickly in dangerous situations and reduce Accident risk, in addition to emergency avoidance decisions, the system can also plan safe routes to avoid risk areas such as congested areas, traffic accidents, bad weather, etc., to improve driving safety and smoothness. The vehicle's intelligent global decision-making engine can comprehensively consider all perception, simulation, and decision-making data to globally optimize the vehicle's behavior. This optimization is not limited to individual scenarios, but covers various situations in the entire driving process, ensuring that the system can adaptively handle complex and changing road environments. Through global optimization, the vehicle can intelligently adjust its driving strategy to avoid traffic accidents, save driving time, and improve traffic smoothness. For example, when congestion or an accident occurs, the system can intelligently adjust the driving route to avoid blocked areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1A schematic diagram of the steps of an intelligent decision support method for vehicle-mounted sensor data fusion according to the present invention; Figure 2 Detailed implementation flow chart of step S1; Figure 3 Detailed implementation flow chart of step S2. DETAILED DESCRIPTION

[0010] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0011] This application example provides an intelligent decision support method and system for vehicle-mounted sensor data fusion. The execution subjects of the intelligent decision support method and system for vehicle-mounted sensor data fusion include but are not limited to: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the system can be regarded as the general computing node of this application, and the data processing platform includes but is not limited to: at least one of an audio and image management system, an information management system, and a cloud data management system.

[0012] See also Figures 1 to 3 The present invention provides an intelligent decision support method for vehicle-mounted sensor data fusion, and the intelligent decision support method for vehicle-mounted sensor data fusion comprises the following steps: Step S1: acquiring vehicle multi-source operating status monitoring data based on vehicle-mounted sensors; performing vehicle state heterogeneous fusion and operating state multi-period change analysis on the vehicle multi-source operating status monitoring data, thereby obtaining vehicle operating state evolution trend characteristics; Step S2: Acquire the vehicle driving recorder monitoring image and vehicle dynamic environment monitoring parameters, perform multi-target visual detection and dynamic operation scene fitting, and construct a vehicle operation scene diagram; Step S3: Modeling the three-dimensional scene point cloud of the vehicle running state scene graph, and rendering the dynamic vehicle trend according to the evolution trend characteristics of the vehicle running state, so as to obtain a dynamic vehicle scene twin model; Step S4: Perform dynamic vehicle operation simulation on the dynamic vehicle scenario twin model, and then perform dynamic emergency risk analysis to generate scenario dynamic emergency risk data; Step S5: Perform dynamic emergency avoidance behavior decision and safe driving path planning on the scene dynamic emergency risk data, thereby generating a visual view of the emergency risk decision; Step S6: Based on the emergency risk decision visualization view, instant perception feedback is performed, and vehicle global self-optimization is performed to build a vehicle intelligent global decision engine.

[0013] In the embodiment of the present invention, refer to Figure 1, is a schematic diagram of the steps of an intelligent decision support method for vehicle-mounted sensor data fusion of the present invention. In this example, the steps of the intelligent decision support method for vehicle-mounted sensor data fusion include: Step S1: acquiring vehicle multi-source operating status monitoring data based on vehicle-mounted sensors; performing vehicle state heterogeneous fusion and operating state multi-period change analysis on the vehicle multi-source operating status monitoring data, thereby obtaining vehicle operating state evolution trend characteristics; In this embodiment, the required types of vehicle-mounted sensors are determined, including GPS, accelerometer, gyroscope, speed sensor, temperature sensor and OBD (on-board diagnostic) system. These sensors can provide multi-dimensional data such as vehicle position, speed, acceleration, steering angle and engine status. Configure the sampling frequency of the sensor. Usually, the sampling frequency of GPS is set to 1Hz, and that of accelerometer and gyroscope is set to 50Hz to ensure that sufficiently fine dynamic state data is obtained. Build a vehicle data acquisition system and use an embedded computing unit (such as Raspberry Pi or Arduino) to collect and store data from various sensors. Ensure that the system has real-time data processing capabilities for subsequent analysis. Use appropriate communication protocols (such as CAN bus or Bluetooth) to aggregate sensor data to the central processing unit to ensure the stability and real-time performance of data transmission. Use data fusion technology (such as weighted averaging, Kalman filter or multi-sensor data fusion algorithm) to process data from different sensors. Kalman filter is particularly suitable for state estimation of dynamic systems and can effectively reduce the impact of noise on data. Set data fusion parameters, such as weighting coefficients and the state transition matrix of the filter, to achieve effective integration of different sensor data. Clean the acquired sensor data to remove outliers and missing values. Use statistical methods (such as Z-score or IQR method) to identify and process outliers in the data to ensure data quality. Standardize data from different time periods to eliminate the impact of different sensor data scales and ensure the accuracy of subsequent analysis. Divide the vehicle operation status monitoring data into different time periods, such as peak hours, off-peak hours, and nighttime hours. The data of each time period should be analyzed independently to identify the operation status characteristics under different conditions. Use time series analysis methods to group the data of each time period to ensure that the data of each group is representative. Use data analysis tools (such as Python's Pandas and NumPy libraries) to perform statistical analysis on the data of each time period and calculate various indicators (such as average speed, acceleration rate of change, and steering frequency). Use visualization tools (such as Matplotlib or Seaborn) to draw the operation status change trend chart of each time period to intuitively display the state changes of the vehicle in different time periods. Based on the extracted operation status data, use machine learning algorithms (such as principal component analysis PCA or linear regression) to extract the evolution trend characteristics of the vehicle operation status. PCA can effectively reduce the dimension and extract the main features. Set the parameters of feature extraction, such as the variance ratio to be retained (usually set to 95%), to ensure that the extracted features are representative.

[0014] Step S2: Acquire the vehicle driving recorder monitoring image and vehicle dynamic environment monitoring parameters, perform multi-target visual detection and dynamic operation scene fitting, and construct a vehicle operation scene diagram; In this embodiment, the sampling frequency of the camera is set, generally 30 frames per second is selected to ensure that every detail in the dynamic scene is captured, the exposure time and gain of the camera are configured to adapt to different lighting conditions (such as day and night), and dynamic environment monitoring parameters are obtained through vehicle-mounted sensors (such as GPS, IMU (inertial measurement unit) and LiDAR). These sensors can provide information about the vehicle's position, speed, acceleration and surrounding obstacles. Ensure that the sampling frequency of the sensor is consistent with that of the driving recorder to maintain data synchronization. Usually, GPS can be set to 1Hz, and IMU and LiDAR can be set to 10Hz or higher to obtain detailed dynamic change information. The acquired monitoring image is preprocessed, including denoising, contrast enhancement and image normalization. Use image processing libraries (such as OpenCV) to implement these steps to ensure that the image quality is suitable for subsequent analysis. Apply edge detection (such as Canny algorithm) and feature point extraction (such as SIFT or ORB) to enhance the visibility of important features in the image. Use deep learning models (such as YOLO, Faster R-CNN or SSD) for multi-target detection to identify objects such as vehicles, pedestrians, traffic signs on the road, etc., select appropriate pre-trained models to accelerate the training process, use labeled data sets (such as COCO or Pascal VOC) for transfer learning during training, set appropriate hyperparameters (such as learning rate, batch size, and number of iterations) to improve detection accuracy, record the location, category, and confidence of each detected object to ensure the traceability of the detection results, build a dynamic running scene model based on the detected target objects and their locations, use 3D modeling tools (such as Blender or Unity) to visualize the detection results as a scene graph, reconstruct the relationship between the vehicle and the surrounding environment, and dynamically fit the scene in combination with the vehicle's dynamic monitoring parameters (such as speed, acceleration, and steering angle) to ensure that the scene model can reflect the real-time dynamic behavior of the vehicle, and record the driving recorder The images, vehicle dynamic monitoring parameters and multi-target detection results are fused to generate a vehicle operation scene graph, which should include information such as vehicles, pedestrians, traffic signs, road structures, etc. The data fusion algorithm (such as Kalman filter) is used to integrate data from different sources to ensure that the generated scene graph has high precision and high real-time performance. The final vehicle operation scene graph is generated, including the identified targets, dynamic environment and vehicle status. The image is clear and the information is complete to facilitate subsequent analysis and decision-making. The scene graph is output in a standard format (such as PNG or JPEG) and the relevant metadata is recorded, including timestamps, detection results and dynamic monitoring parameters.

[0015] Step S3: Modeling the three-dimensional scene point cloud of the vehicle running state scene graph, and rendering the dynamic vehicle trend according to the evolution trend characteristics of the vehicle running state, so as to obtain a dynamic vehicle scene twin model; In this embodiment, the key elements in the scene, including vehicles, pedestrians, traffic signs, and road structures, are extracted by combining the vehicle running scene graph generated in the previous stage. Ensure that the geometric information and position coordinates of all elements are clear. The original point cloud data obtained by sensors such as LiDAR or depth cameras are combined with the scene graph to form a more complete three-dimensional scene foundation. The integrated data is processed using a point cloud processing library (such as PCL - Point Cloud Library) to generate a three-dimensional point cloud. Ensure that each point cloud data point contains spatial coordinates (x, y, z) and other attributes (such as color, intensity, etc.). Filter and denoise the point cloud data to improve the quality of the point cloud. Apply voxel grid filter (Voxel GridFilter) or statistical outlier removal (Statistical Outlier Removal) algorithm to ensure that the generated point cloud data has high accuracy. Use three-dimensional modeling software (such as Blender, Unity, or Maya) to convert the generated point cloud data into a three-dimensional model. During the modeling process, ensure that each element in the scene can accurately reflect its shape and position in three-dimensional space. Use mesh reconstruction algorithms (such as Poisson Surface Reconstruction or Greedy Triangulation) to convert point cloud data into continuous 3D meshes to ensure the smoothness and realism of the model. Add materials and textures to the generated 3D model. According to the color and lighting conditions in the actual scene, select appropriate texture mapping technology (such as UV mapping) to color the model to ensure the visual effect of the model is realistic. In this process, record material properties (such as reflectivity, glossiness, etc.) for subsequent dynamic rendering. Based on the vehicle running state evolution trend characteristics extracted in the previous stage, analyze the dynamic behavior of the vehicle in different time periods, such as acceleration, braking, and steering. Convert these dynamic behaviors into animation parameters for rendering in the 3D model. Use time series analysis methods (such as ARIMA models) to model the dynamic behavior of the vehicle and extract key dynamic features and trend changes. In the 3D modeling software, use animation tools (such as skeletal animation or shape keyframes) to add dynamic behavior to the vehicle model. According to the extracted trend characteristics, set the state changes of the vehicle at different time points to ensure that the dynamic effect of the rendering is natural and smooth. Apply a physics engine (such as Unity's PhysX or Blender's Bullet) to simulate the real dynamic behavior of the vehicle in the scene, including collision detection and physical reactions, to enhance the realism of the model.

[0016] Step S4: Perform dynamic vehicle operation simulation on the dynamic vehicle scenario twin model, and then perform dynamic emergency risk analysis to generate scenario dynamic emergency risk data; In this embodiment, a dynamic vehicle operation simulation environment is built using professional simulation software (such as CARLA, SUMO or MATLAB / Simulink). Ensure that the environment can simulate real traffic scenes, including roads, traffic signals, pedestrians and other vehicles. Configure environmental parameters such as road type (urban roads, highways), traffic density (such as low, medium, and high density) and weather conditions (sunny, rainy, and night) to facilitate subsequent simulation tests. Use a vehicle dynamics model (such as a Bike model or a four-wheel model) to define the motion characteristics of the vehicle. Set model parameters, including vehicle mass, tire characteristics, turning radius, and acceleration limits to ensure that the simulated vehicle behavior is real and reliable. Perform model verification and ensure that the response of the simulation model under different conditions is consistent with the actual situation by comparing with the actual vehicle test data. Start the simulation process, set the simulation time and step size (for example, the simulation time is 60 seconds and the time step size is 0.1 seconds) to ensure that sufficient dynamic data is obtained. During the simulation process, record the vehicle's state data, including information such as position, speed, acceleration, and steering angle, for subsequent analysis. Based on the characteristics of dynamic vehicle operation, define emergency risk indicators, such as collision risk, frequency of sudden braking, sharpness of turning, and proximity of pedestrians. These indicators can help identify potential risk scenarios. Set risk assessment thresholds. For example, the threshold for collision risk is set to 0.5, and when the frequency of sudden braking exceeds a certain value, it is considered high risk. Use rule-based models or machine learning models (such as decision trees or random forests) for dynamic emergency risk assessment. Use the vehicle status data recorded during the simulation as input to evaluate the risk level at each time step. Set model parameters, such as the ratio of training set to test set (usually 80:20), to ensure that the training and evaluation of the model are representative. During the simulation process, calculate the risk indicator for each time step in real time and compare it with the preset threshold to generate dynamic emergency risk data. Record the risk level, corresponding vehicle status and environmental information at each time step, and generate a structured risk data report. Perform statistical analysis on the generated dynamic emergency risk data, calculate the average, maximum and change trend of each risk indicator, and identify high-risk periods and scenarios. Use time series analysis methods (such as sliding average or exponential smoothing) to smooth the risk data to help identify potential risk patterns.

[0017] Step S5: Perform dynamic emergency avoidance behavior decision and safe driving path planning on the scene dynamic emergency risk data, thereby generating a visual view of the emergency risk decision; In this embodiment, key parameters are extracted from the acquired dynamic emergency risk data, including accident risk level, vehicle status (position, speed, acceleration) and environmental information (road conditions, pedestrian distribution), to ensure that these data can provide a basis for subsequent decision-making processes, set a data time window, for example, 1 second as a time step, to facilitate real-time decision-making and path planning, and classify the dynamic emergency risk data according to predefined risk thresholds, for example, the risk level is divided into three levels of "low", "medium" and "high", to facilitate quick judgment when making decisions, use statistical analysis methods (such as standard deviation and variance analysis) to identify outliers in risk data, ensure that decisions are based on reliable data, and adopt a rule-based decision model. Or machine learning algorithms (such as reinforcement learning) to define emergency avoidance behaviors, set decision rules, for example, when the risk level is "high", priority is given to deceleration or steering avoidance, set the input parameters of the decision model, including the current risk level, vehicle status, and surrounding environment information (such as obstacle location and type), implement emergency avoidance decisions in the simulation environment, and perform corresponding avoidance behaviors such as deceleration, steering or parking based on real-time dynamic risk data. Record the results of decision execution, including avoidance success rate, collision occurrence and emergency response time, etc., for subsequent analysis and optimization, and select appropriate path planning algorithms, such as A algorithm, Dijkstra algorithm or RRT (Rapidly-exploring Random Tree) algorithm is used to realize safe driving path planning. Algorithm A is more efficient in complex environments, while Dijkstra algorithm is suitable for deterministic scenarios. It sets the goals of path planning, including the shortest path, the safest path, and avoiding high-risk areas. According to the current vehicle position and target position, a safe path is generated using the selected path planning algorithm to ensure that the path avoids high-risk areas and considers traffic signals and other dynamic obstacles. During the path generation process, the distance, estimated time, and risk assessment of each path are recorded to facilitate the subsequent selection of the best path.

[0018] Step S6: Based on the emergency risk decision visualization view, instant perception feedback is performed, and vehicle global self-optimization is performed to build a vehicle intelligent global decision engine.

[0019] In this embodiment, key data (such as risk level, vehicle status, and path planning information) in the emergency risk decision visualization view are integrated into the vehicle's perception system to ensure that the data is updated in real time for quick decision-making. The update frequency of the perception data is set, such as once per second, to ensure the timeliness and accuracy of the acquired data. An instant feedback mechanism is developed to ensure that the vehicle can make real-time adjustments according to the risk decision visualization view during driving. For example, when the risk level increases, the vehicle should automatically slow down or change the driving direction. Feedback rules are set. For example, if the risk level exceeds a set threshold (such as "high"), the corresponding avoidance action is triggered, and the response time and execution effect of each feedback are recorded. The goal of global self-optimization is determined, including improving driving safety, reducing emergency response time, and optimizing driving paths. It is ensured that each goal can be quantitatively evaluated through a data-driven method. Optimization indicators, such as the number of collisions, emergency braking frequency, and driving time, are set to facilitate subsequent evaluation of the optimization effect. The self-optimization model is trained using historical data and real-time perception feedback data. A suitable machine learning algorithm (such as reinforcement learning or deep learning) is selected to optimize the decision-making process. Training parameters, such as learning rate, discount factor, and training rounds, are set to ensure that the model can effectively learn and self-adjust. Usually Select 80% of the data for training and 20% of the data for testing and verification. Design the architecture of the intelligent global decision-making engine, including the data input module, decision optimization module and execution module, to ensure efficient collaboration between the modules to achieve real-time decision-making. Determine the working mode of the engine, such as real-time decision-making, offline learning and self-optimization, to improve the flexibility and adaptability of the system. In the decision optimization module, apply optimization algorithms (such as genetic algorithms, particle swarm optimization, etc.) to analyze real-time data and generate optimal decision-making solutions. These algorithms can find the optimal solution in a complex environment. Set the parameters of the optimization algorithm, such as population size, number of iterations and fitness. function to ensure that the algorithm can find the optimal solution within a reasonable time, monitor and evaluate the output of the intelligent global decision-making engine in real time to ensure that it can effectively execute decisions in a dynamic environment, record the execution of each decision, including success rate and response time, use statistical analysis methods (such as A / B testing) to compare the decision effects before and after optimization, evaluate the effectiveness of system improvements, adjust and optimize the engine according to the verification results, including modifying model parameters, optimizing algorithms and feedback rules, ensure that the engine can continuously learn and adapt to new traffic environments, record the process and experimental parameters of each optimization, and provide a basis for subsequent engine iterations and upgrades.

[0020] In this embodiment, refer to Figure 2 , is a flowchart of detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Step S11: acquiring vehicle multi-source operating status monitoring data based on vehicle-mounted sensors; Step S12: performing vehicle state heterogeneous fusion on vehicle multi-source operation state monitoring data, thereby generating multi-dimensional vehicle operation state features; Step S13: performing a multi-period change analysis on the multi-dimensional running state characteristics of the vehicle to generate the running state characteristics of the vehicle in multiple time periods; Step S14: Evolving the state change trend of the vehicle running state characteristics in multiple time periods, thereby obtaining the vehicle running state evolution trend characteristics.

[0021] In this embodiment, a data acquisition system is configured to ensure that the data of all sensors can be concentrated in a data processing center. Use an on-board computer or data logger to collect data. Set the data acquisition frequency, for example, collect data once every second to ensure real-time performance. Adjust the acquisition frequency according to the driving status of the vehicle to optimize data storage and processing. Use a database (such as MySQL, MongoDB) to store the collected data to ensure the security and integrity of the data. Establish a corresponding table for each sensor data to record information such as timestamp, sensor type, data value, etc. Back up data regularly to ensure that it can be quickly restored when the data is lost or damaged. Clean the collected multi-source data to remove duplicate data and outliers. Use Python's Pandas library for data processing to ensure data accuracy and consistency. Standardize data from different sources to ensure unit consistency. For example, unify the speed to kilometers per hour (km / h) and the time to UTC time format. Synchronize the data of different sensors in time to ensure that the data in the same time period can correspond correctly. Use interpolation to fill in missing data to ensure that each time point has complete status information. Select appropriate data fusion methods, such as weighted average method, Kalman filter or deep learning model (such as LSTM), to fuse multi-source data into unified state features. Consider the importance of each sensor data and set the corresponding weight. Use Kalman filter to achieve real-time state estimation, which is particularly suitable for state fusion of dynamic systems and can effectively reduce the impact of noise. Extract multi-dimensional operating state features from the fused data, including key indicators such as speed, acceleration, fuel consumption, engine temperature, etc. Set the standard and method of feature extraction according to actual application requirements. Generate a comprehensive state feature data set to record the vehicle status in different time periods and provide basic data for subsequent analysis. Divide the data into different time periods according to the vehicle's operating characteristics and analysis requirements, such as by hour, day or specific trip. Ensure that the data in each time period can reflect the vehicle's operating status. Set the length of the time period, such as one hour or every 5 minutes as an analysis unit to ensure the meticulousness of the analysis. Aggregate the data in each time period and calculate the statistical values ​​of each operating state feature, such as the mean, maximum, minimum and standard deviation. Use aggregation functions (such as mean, max, min) to process the data. Record the running characteristics of each time period and store them in a new data table for subsequent analysis. Use appropriate trend analysis methods, such as time series analysis, regression analysis, or data mining techniques (such as cluster analysis) to identify the changing trends of vehicle running characteristics in different time periods. Use Python's statsmodels library to perform time series analysis to identify and predict trend changes. Visualize the analysis results and use charts (such as line charts and bar charts) to show the changing trends of vehicle running characteristics for intuitive understanding. Make sure the charts are clear and can effectively convey information.Record the significance of the change trend and identify the key factors that affect the vehicle's operating status, such as traffic conditions, weather changes, etc. Select a suitable evolution model, such as a Markov model, a state transition model, or a deep learning model (such as LSTM) to analyze the evolution trend of the vehicle's operating status. Select the optimal model based on the data characteristics and analysis requirements. Configure the model parameters, set the state transition probability and initial state, and ensure that the model adapts to the actual data. Use the collected multi-period vehicle operating state feature data to train the selected evolution model. Ensure the diversity and representativeness of the training data to improve the generalization ability of the model. Perform cross-validation to evaluate the accuracy and reliability of the model. Adjust the model parameters based on the validation results to optimize the model performance. Based on the trained model, analyze the evolution trend of the vehicle's operating status in different time periods. Record the transition probability and changes of each state and identify the key factors that affect the evolution. Generate an evolution trend report, including the operating state changes and influencing factors in each time period, to provide a basis for subsequent decision-making.

[0022] In this embodiment, the specific steps of step S12 are: The vehicle multi-source operating status monitoring data includes vehicle dynamics data, engine monitoring data, battery status monitoring data, tire pressure sensor parameters and IMU real-time monitoring parameters; Perform vehicle speed change analysis on vehicle dynamics data to obtain vehicle time-series speed change characteristics; Perform engine temperature change identification on engine monitoring data and extract engine temperature change data; Perform discrete temperature variation fitting on the engine temperature variation data to construct the engine temperature variation curve; Perform real-time load analysis on battery status monitoring data to obtain real-time battery load status data; Perform tire pressure stability identification on tire pressure sensing parameters to generate vehicle tire pressure operation balance characteristics; Carry out body posture fluctuation detection on the IMU real-time monitoring parameters to generate body posture fluctuation data; The vehicle state heterogeneous fusion is performed on the vehicle's time-series speed change characteristics, engine temperature change curve, battery real-time load status data and body posture fluctuation data to generate multi-dimensional vehicle operation state characteristics.

[0023] In this embodiment, the basic statistical characteristics of the vehicle speed are calculated, including the average speed, the highest speed, the lowest speed and the speed fluctuation range. The Python Pandas library is used for data analysis. According to the time series data, the speed change rate is calculated, the acceleration and deceleration periods are identified, and a speed change graph (such as a line graph) is generated to intuitively display the speed change characteristics. A speed change curve graph is drawn to indicate the speed changes under different driving conditions, such as starting, accelerating, cruising and decelerating. The analysis results are recorded and a detailed speed change report is generated, including speed characteristics and change trends, to provide data support for subsequent analysis. The engine temperature data is obtained from the vehicle-mounted OBD system or the engine control unit (ECU), and the timestamp and temperature value (such as Celsius), set the data collection frequency to every second to obtain real-time temperature changes, clean the collected temperature data, remove outliers and noise, identify and eliminate unreasonable temperature values ​​(such as values ​​below the ambient temperature or above the normal operating temperature of the engine), ensure the uniformity of timestamps, and keep them consistent with vehicle speed data for subsequent analysis, calculate the basic statistical characteristics of engine temperature, including maximum temperature, minimum temperature, average temperature and temperature fluctuation range, use Pandas library for data processing and analysis, identify key temperature change events during engine operation, such as temperature changes during startup, temperature rise during acceleration and stable temperature during idling, draw temperature change curves, and mark the temperatures under different working conditions Changes, such as idling, acceleration and high-speed driving, generate a detailed temperature change identification report, including temperature change characteristics and trends, to provide a basis for subsequent analysis, discretize the temperature change data, calculate the average, maximum and minimum values ​​of the temperature data according to the time period (such as 5 seconds, 10 seconds), generate a discrete data set, ensure that the discretized data can effectively reflect the temperature change trend, avoid information loss caused by over-simplification, select a suitable fitting method, such as polynomial fitting, spline interpolation or linear regression, fit the discrete temperature data, use the SciPy library in Python for fitting, select a suitable fitting function and parameters, use the selected fitting method to fit the discretized temperature data, and obtain the engine temperature change curve, record the parameters of the fitting model and the goodness of fit (such as R² value), draw a comparison chart of the fitting curve and the original data through visualization tools (such as Matplotlib) to intuitively display the fitting effect, obtain battery status monitoring data from the battery management system (BMS), including voltage, current, temperature and charging status, etc., and set the acquisition frequency to every second to ensure that the data can reflect the performance and status of the battery under different working conditions. Clean the original data, remove outliers and noise, and ensure the accuracy of the data. Use Python's Pandas library for data processing, record the battery load status at each time point, generate a complete data set, and calculate the real-time load (power) of the battery based on the current and voltage data. Use the formula P = V * I, where P is power,V is voltage, I is current, calculate the average value, maximum value and fluctuation range of the load, analyze the load changes of the battery under different working conditions, draw a load change curve, mark the load changes of the battery under different working conditions, such as acceleration, idling and braking, generate a battery load status analysis report, including real-time load characteristics and change trends, provide a basis for subsequent decision-making, obtain real-time tire pressure data of each tire from the tire pressure monitoring system (TPMS), record the timestamp and tire pressure value (such as kPa) of each tire, ensure that the data collection frequency is consistent with other sensors, and facilitate subsequent analysis, clean the collected tire pressure data, remove unreasonable tire pressure values ​​(such as negative values ​​or extremely high values), ensure the validity of the data, and store the tire pressure data. Standardization is performed to ensure data consistency. Appropriate stationarity analysis methods are selected, such as standard deviation calculation, root mean square error (RMSE) or stationarity test (such as ADF test). These methods can effectively identify the volatility of tire pressure. An analysis time window is set. For example, the stationarity index of tire pressure is calculated every 5 minutes. The mean, standard deviation and RMSE of tire pressure in each time window are calculated to identify the stationarity of tire pressure. Vehicle body posture data, including acceleration, angular velocity and tilt angle, are obtained from the inertial measurement unit (IMU). The data collection frequency is set to every second to ensure that the data can accurately reflect the changes in the vehicle's posture during driving. The original posture data is cleaned to remove outliers and noise to ensure data accuracy. In order to ensure the validity of the data, use filters (such as low-pass filters) to smooth the data and reduce the impact of noise. Standardize the data to ensure that the units of various parameters are consistent. Calculate the fluctuation characteristics of the vehicle posture, including the mean, standard deviation, maximum and minimum values ​​of the acceleration. Identify the stability and dynamic changes of the vehicle during driving. Identify key events, such as posture changes during rapid turns, rapid acceleration or deceleration. Draw a posture fluctuation change graph to indicate the posture changes under different working conditions, such as driving, turning or braking. Integrate the vehicle speed, engine temperature, battery load and body posture data into a unified data framework to ensure consistent timestamps for subsequent analysis. Standardize data from different sources. , to ensure the consistency and comparability of data, select appropriate data fusion methods, such as weighted average, data interpolation or machine learning methods (such as ensemble learning) to achieve the fusion of multi-source data, determine the weight of each data source, consider its importance in the vehicle status, fuse the multi-source data according to the selected method, generate comprehensive multi-dimensional vehicle operation status characteristics, record the fusion process and calculation formula of each feature, ensure transparency and traceability, visualize the fused data, and display the comprehensive operation status of the vehicle, such as using radar charts or multi-dimensional line charts, to generate a detailed report on the multi-dimensional operation status characteristics of the vehicle, including the fused features and their physical meanings, to provide a basis for subsequent decision-making and optimization.

[0024] In this embodiment, refer to Figure 3 , is a flowchart of detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Step S21: Acquire the vehicle driving recorder monitoring image and vehicle dynamic environment monitoring parameters; Step S22: Analyzing the current environment change characteristics based on the vehicle dynamic environment monitoring parameters to obtain the current vehicle operating environment change characteristics; Step S23: performing multi-target visual detection on the vehicle driving recorder monitoring image and marking each scene target; Step S24: performing deep semantic analysis on each scene object, thereby generating semantic features of each scene object; Step S25: performing real-time operation scenario analysis based on the semantic features of each scenario target to generate vehicle real-time operation scenario data; Step S26: Perform dynamic operation scenario fitting on the real-time vehicle operation scenario data based on the current vehicle operation environment change characteristics to construct a vehicle operation state scenario diagram.

[0025] In this embodiment, the vehicle is equipped with dynamic environment sensors, including a GPS module, an accelerometer, a temperature sensor, and a light sensor. These sensors will record parameters such as the vehicle's speed, position, acceleration, ambient temperature, and light intensity in real time. The data acquisition frequency of the sensor is set, usually 1 second, to ensure the real-time and accuracy of the data, and the data is stored in the on-board computer. The video data of the driving recorder and the dynamic environment monitoring parameters are stored in the hard disk of the on-board computer to ensure the security and integrity of the data. The timestamp of each video clip is recorded for synchronization during subsequent analysis. The data structure is designed to ensure that the video data and the monitoring parameters are easily associated and retrieved, which is convenient for subsequent data processing. The GPS, temperature, light intensity, acceleration, and other data are extracted from the dynamic environment monitoring system to form a data set containing multiple features, ensuring that the timestamp of the data is consistent with the driving recorder. The video data of the vehicle recorder is consistent for subsequent analysis. Data cleaning is performed to remove outliers and missing data to ensure data integrity and accuracy. The key features of the analysis are determined, such as speed change rate, temperature change amplitude, acceleration change, etc. According to actual application requirements, the characteristic indicators of the analysis are set, and each feature is standardized to eliminate the impact of different dimensions and ensure the comparability between features. Statistical methods are used to calculate the changes of each feature within a certain time window, such as mean, standard deviation and change rate. The time window is selected as 5 minutes to identify more obvious environmental changes. The environmental feature change graph is drawn through visualization tools (such as Matplotlib) to show the environmental change trend in different time periods. The analysis results are recorded and the environmental change feature report is generated, including the change trend and influencing factors of environmental features, which provides a basis for subsequent scene analysis.

[0026] Identify key environmental change events, such as high temperature, strong light, or drastic acceleration changes, and analyze their impact on vehicle operation. Extract each frame of image from the video data obtained from the dashcam to ensure that the image quality is good and suitable for subsequent analysis. Select to extract 1 frame of image every 1 second for processing. Preprocess the extracted images, such as resizing, denoising, and enhancing contrast, to improve the accuracy of subsequent visual inspection. Select a deep learning target detection model, such as YOLO (You Only Look Once) or Faster R-CNN, these models perform well in multi-target detection and can recognize multiple objects in an image in real time, prepare a dataset, ensure that the training set contains a variety of scenes and targets (such as pedestrians, vehicles, traffic signs, etc.), and annotate them, use the annotated dataset to train the selected target detection model, set appropriate hyperparameters such as learning rate, batch size, and number of iterations, use deep learning frameworks such as TensorFlow or PyTorch for training, perform model validation, ensure that the model performs well on the test set and achieves the expected detection accuracy, apply the trained model to the video data of the dashcam, identify and mark the targets in each frame of the image, record the category and location coordinates of each target, post-process the detection results, ensure accurate marking and eliminate redundant detections, draw a detection result map, mark the identified targets with bounding boxes and annotate category information, generate a target detection report, determine the semantic features that need to be extracted, including the target's category, behavior, state, and relationship with the surrounding environment, etc., ensure that these features can fully describe each target in the scene, prepare a relevant semantic database to provide detailed background information for each target, and select a suitable deep learning model for semantic Parsing, such as BERT or GPT-3, these models can process natural language and generate rich semantic descriptions, prepare training datasets containing target features, ensure that the datasets cover a variety of scenarios and target information, use the prepared datasets to train the selected models, set appropriate hyperparameters, ensure that the models can accurately generate semantic features for each target, perform model validation, test the performance of the models on new data, ensure the accuracy and reliability of semantic parsing, apply the trained models to the target detection results, extract the semantic features of each target, record the category, motion state, relationship with the environment and other information of each target, associate the extracted semantic features with the target detection results, generate a target dataset containing semantic information, generate a detailed semantic feature report, record the semantic information of each target, provide support for subsequent scene analysis, analyze the changes of semantic features in different scenarios, identify the key factors affecting scene understanding, integrate the target detection results with the deep semantic parsing results, form a scene dataset containing multiple features, ensure that the detection information and semantic features of each target can correspond one to one, record the scene data at each time point for real-time analysis, and build a real-time analysis system.Use stream processing frameworks (such as Apache Kafka and Apache Flink) to process vehicle operation scenario data, ensure the real-time data processing and analysis, set analysis indicators such as traffic density, pedestrian activity, and traffic signal status for comprehensive evaluation, conduct real-time scenario analysis based on the integrated scenario data, evaluate the characteristics and changes of the current traffic environment, use statistical analysis methods to calculate indicators such as traffic density, number of pedestrians, and vehicle speed, identify key events and abnormal situations, such as sudden traffic accidents and traffic signal changes, improve the accuracy of scenario analysis, generate real-time operation scenario analysis reports, record current environmental characteristics, traffic conditions, and potential risks, provide support for subsequent decision-making, and use visualization tools to display real-time scenario analysis results, such as using heat maps to display traffic density, improve the intuitiveness of decision-making, and select appropriate dynamic scenario fitting models. Models, such as state-space-based models, Kalman filters, or deep learning models (such as LSTM), are used to adapt to the characteristics of real-time data. Input parameters of the fitting model, including vehicle speed, surrounding environment, and target features, are determined to ensure that the model can fully reflect the current scene state. The selected model is trained using historical data to ensure that the model can accurately fit the changing characteristics of the vehicle operation scene. Appropriate hyperparameters are set to optimize model performance. Model validation is performed to test the performance of the model on new data to ensure its stability and reliability. Real-time operation scene data is input into the trained model to perform dynamic scene fitting, generate the vehicle's operation scene graph, record key parameters and states during the fitting process, output the fitting results, and generate a comprehensive scene graph containing vehicle status, environmental features, and target information.

[0027] In this embodiment, the detailed implementation steps of step S3 include: Step S31: accurately calculating the spatial position of each scene object to obtain the spatial position coordinates of each scene object; Step S32: performing a three-dimensional road structure analysis on the vehicle running state scene graph, thereby generating three-dimensional structure data of the running scene; Step S33: performing three-dimensional scene point cloud modeling on the three-dimensional structure data of the running scene based on the spatial position coordinates of each scene object, and constructing a real-time three-dimensional scene model of the vehicle; Step S34: Perform dynamic vehicle trend rendering on the real-time three-dimensional scene model of the vehicle according to the evolution trend characteristics of the vehicle operation state, so as to obtain a dynamic vehicle scene twin model.

[0028] In this embodiment, sensors such as GPS, IMU (inertial measurement unit) and LiDAR are used to obtain the real-time position and attitude information of the vehicle. GPS is used to obtain the geographic location, IMU is used to obtain the acceleration and angular velocity of the vehicle, and LiDAR is used to obtain three-dimensional point cloud data of the surrounding environment. Ensure that the acquisition frequency of all sensors is consistent, usually set to 1Hz, to keep the data synchronized. Clean the collected data to remove outliers and noise to ensure the accuracy of the data. Use filters (such as Kalman filters) to smooth GPS and IMU data to reduce the impact of instantaneous fluctuations. Integrate all data into a unified data framework and mark timestamps for subsequent analysis. Convert GPS coordinates to coordinates in the vehicle coordinate system for subsequent spatial position calculations. Use the coordinate conversion formula to convert longitude and latitude to Cartesian coordinate system (X, Y, Z). Combined with IMU data, calculate the vehicle's attitude (pitch angle, yaw angle, and roll angle) in three-dimensional space to ensure the accuracy of the spatial position. For each detected scene target (such as other vehicles, pedestrians, traffic signs, etc.), calculate its position coordinates in three-dimensional space based on its pixel position in the image, combined with the camera's intrinsic and extrinsic parameters. Use three-dimensional reconstruction technology (such as stereo vision or monocular depth estimation) to improve the accuracy of target position calculation and ensure that the spatial position coordinates of each target are accurate. Record the spatial position coordinates of each target and store them together with the target's category and other feature data for subsequent analysis. Obtain point cloud data of the road and surrounding environment from the LiDAR sensor, which contains the shape, width, and longitudinal characteristics of the road. Ensure that the density of the point cloud data is high enough for subsequent three-dimensional structural analysis. Filter and denoise the acquired point cloud data to remove unnecessary background noise and outliers. Use statistical outlier removal methods (such as StatisticalOutlierRemoval in the PCL library) for processing. Downsample the point cloud data to improve processing efficiency while maintaining the integrity of structural features. Use three-dimensional modeling technology (such as the RANSAC algorithm) to extract the planar features of the road and identify the boundaries and center lines of the road. This method can effectively extract straight line and plane features from point clouds. Calculate geometric features such as road width, slope and curvature to analyze the three-dimensional structure of the road. Use the extracted features to build a three-dimensional model of the road and combine it with the vehicle's operating status to generate a complete three-dimensional structural data of the road scene. Use three-dimensional modeling software such as Meshlab or Blender to build the model. Record the analysis results, including the geometric features and three-dimensional structural information of the road, to provide support for subsequent scene modeling. Integrate the spatial position coordinates of each scene target with the previously generated three-dimensional structure data of the road to ensure that all elements are processed in the same coordinate system. Record the spatial information and road features of each target for subsequent modeling.Combine LiDAR point cloud data with the spatial coordinates of the target to form a comprehensive point cloud dataset that represents the 3D structure of the entire scene. Use a point cloud processing library (such as PCL) to merge and process point cloud data to ensure data consistency and integrity. Select appropriate point cloud modeling techniques, such as voxel grid or surface reconstruction techniques (such as Poisson reconstruction) to generate high-quality 3D scene models. Set model parameters, such as voxel size and reconstruction accuracy, to ensure that the generated 3D model meets the requirements of the actual scene. Generate a real-time 3D scene model of the vehicle based on the integrated point cloud data and target coordinates. Use software such as MeshLab or CloudCompare for point cloud reconstruction. Record the properties of each point cloud, such as color and reflectivity, to improve the realism of the scene model. Generate a 3D scene model and its feature report, including the accuracy and level of detail of the model. Use visualization tools to display the generated 3D model to support subsequent analysis. Based on historical data and real-time data, extract the vehicle's running state evolution trend characteristics, such as speed changes, acceleration changes, steering angles, and driving paths. Record key features in each time period for dynamic trend analysis. Apply the extracted dynamic features to the generated 3D scene model to build a dynamic vehicle trend model. Use a game engine (such as Unity or Unreal Engine) for dynamic rendering to achieve real-time scene display. Set the vehicle's motion trajectory and dynamic behavior to reflect the real-time dynamic changes of the vehicle in the model. Optimize the rendering effects, including lighting effects, material textures, and shadow effects, to improve the realism and immersion of the scene. Record performance indicators during the rendering process, such as frame rate and rendering latency, to ensure that the model can run smoothly in a real-time environment. Generate a scene twin model that includes the dynamic behavior of the vehicle, and record the status and parameters of each scene for subsequent analysis and optimization. Generate a dynamic scene twin model report, including the dynamic characteristics and performance of the model, to provide a basis for decision-making in the intelligent transportation system.

[0029] In this embodiment, step S4 includes the following steps: Step S41: performing dynamic vehicle operation simulation on the dynamic vehicle scenario twin model to generate vehicle operation simulation response data; Step S42: pre-identify the road condition of the dynamic vehicle scene twin model and extract the road condition data ahead; Step S43: predicting the current vehicle driving situation based on the vehicle operation simulation response data based on the front road condition data to generate vehicle driving situation prediction data; Step S44: performing road vehicle operation trend prediction on the vehicle operation simulation response data to obtain road vehicle trend prediction data; Step S45: calculating the probability of pedestrian crossing the road based on the vehicle operation simulation response data, thereby generating the probability of pedestrian crossing the road; Step S46: Perform dynamic emergency risk analysis based on the road vehicle trend prediction data, the road pedestrian crossing probability and the vehicle driving situation prediction data, so as to generate scene dynamic emergency risk data.

[0030] In this embodiment, the previously constructed dynamic vehicle scenario twin model is used to ensure that the model accurately reflects the physical characteristics and environmental factors of the vehicle. The model should include vehicle dynamics, the surrounding environment, and the behavior of other traffic participants. Simulation parameters, such as the vehicle's initial speed, acceleration, steering angle, etc., are configured to facilitate multi-scenario simulation testing. A suitable simulation platform, such as SUMO (Simulation of Urban MObility) or MATLAB / Simulink, is selected to perform dynamic vehicle operation simulation. These tools can simulate complex traffic conditions and vehicle behaviors. According to the simulation objectives, the simulation duration and time step are set. Usually, a time step of 1 second is selected to improve the simulation accuracy. The simulation is started. According to the preset vehicle initial state and environmental conditions, the vehicle's motion trajectory and state changes are calculated in real time. The vehicle position, speed, acceleration, and other information at each time step are recorded. A physical engine (such as Bullet or PhysX) is used to simulate the vehicle's dynamic behavior, including acceleration, braking, steering, etc., to ensure the authenticity of the simulation results. During the simulation process, vehicle operation simulation response data is generated in real time, including the vehicle's speed curve, acceleration curve, trajectory data, etc. Record the timestamp and status information of the data to ensure the integrity of the data, perform data visualization, and use graphical tools (such as Matplotlib or Plotly) to display the simulation results for easy analysis and evaluation. Use computer vision technology to preprocess the images obtained by the driving recorder to extract information such as road boundaries, lane lines, and traffic signs. Use edge detection (such as the Canny algorithm) and Hough transform for linear feature extraction. Combine LiDAR point cloud data to obtain the three-dimensional structural information of the road to ensure a comprehensive understanding of the road conditions. Fuse data from different sensors (such as cameras, LiDAR, GPS) to generate a comprehensive road condition model. Use Kalman filters for data fusion to ensure data accuracy and consistency.

[0031] Analyze the extracted road features, identify changes in road conditions, such as good road surface, potholes, slippery road surface, and traffic jams, use machine learning algorithms (such as SVM or decision trees) to classify the state, record the characteristic data of each road state, including road width, lighting conditions, road surface type, etc., for subsequent analysis, generate road condition data reports, record the state information of the road ahead, and provide a basis for vehicle driving decisions. Use visualization tools to display road condition information for intuitive understanding, integrate vehicle operation simulation response data with road condition data ahead, ensure that the vehicle state at each time step corresponds to the road condition information, record timestamps for analysis, and prepare input data for the prediction model, including vehicle speed, acceleration, and Speed, steering angle, and characteristic data of the road conditions ahead, select a suitable prediction model, such as a time series prediction model (such as ARIMA) or a machine learning model (such as LSTM) for driving situation prediction. LSTM can process time series data and is suitable for vehicle dynamic prediction. Set the model's hyperparameters, such as learning rate, batch size, and number of training rounds, to ensure the performance of the model during the training process. Use historical data to train the selected prediction model, verify the prediction accuracy of the model on the test set, adjust the hyperparameters to optimize the model performance, perform cross-validation, ensure the generalization ability of the model, test the stability of the model on different data sets, input the integrated data into the trained model, predict the current vehicle driving situation, and record each The driving situation prediction results of each time step, including information such as speed, acceleration and driving direction, generate a vehicle driving situation prediction data report, record the prediction results and their accuracy analysis in detail, integrate the vehicle operation simulation response data with the road condition data, ensure that the vehicle state at each time step corresponds to the road trend information, prepare the input data of the prediction model, including vehicle operation parameters (speed, acceleration) and road characteristics, select a suitable trend prediction model, such as linear regression, support vector regression (SVR) or deep learning model (such as RNN) for trend prediction, linear regression is suitable for simple trend analysis, while SVR and RNN are suitable for complex trend analysis, set the model's hyperparameters to ensure the performance of the model during training, Use historical data to train the selected prediction model, verify the prediction accuracy of the model on the test set, record the performance indicators of each model, such as mean square error (MSE) and R² value, perform cross-validation to ensure the generalization ability of the model, and test the stability of the model on different data sets. Input the integrated data into the trained model to predict the road vehicle operation trend, record the prediction results of each time step, including information such as vehicle density, flow and speed, generate a road vehicle operation trend prediction data report, record the prediction results and their accuracy analysis in detail, integrate the vehicle operation simulation response data with the road condition data, ensure that the vehicle status at each time step corresponds to the pedestrian activity information, prepare the input data of the pedestrian crossing probability model,Including vehicle speed, distance, road characteristics ahead and pedestrian location data, select appropriate pedestrian crossing probability calculation models, such as Logistic regression, decision tree or deep learning model (such as CNN) for probability calculation, these models can effectively predict the probability of pedestrian crossing, set the model's hyperparameters, ensure the performance of the model during the training process, use historical data to train the selected probability model, verify the prediction accuracy of the model on the test set, record the performance indicators of each model, such as accuracy and F1-score, perform cross-validation to ensure the generalization ability of the model, and test the stability of the model on different data sets, input the integrated data into the trained model, calculate the probability of pedestrian crossing, record the probability prediction results of each time step, generate a road pedestrian crossing probability data report, record the probability calculation results and its accuracy analysis in detail, integrate the road vehicle trend prediction data, pedestrian crossing probability and vehicle driving situation prediction data, and ensure that the emergency risk analysis of each time step can be based on comprehensive data support. ,

[0032] Prepare input data for the dynamic emergency risk analysis model, including information such as vehicle density, pedestrian probability, speed and acceleration, select appropriate risk analysis models such as Bayesian networks, decision trees or deep learning models (such as LSTM) for emergency risk analysis, which can comprehensively evaluate various risk factors, set the model's hyperparameters, ensure the performance of the model during the training process, use historical data to train the selected risk analysis model, verify the model's prediction accuracy on the test set, record the performance indicators of each model, such as accuracy and recall, perform cross-validation to ensure the generalization ability of the model, and test the stability of the model on different data sets, input the integrated data into the trained model, conduct dynamic emergency risk analysis, record the risk score and risk level for each time step, generate a scenario dynamic emergency risk data report, and record the risk analysis results and their accuracy analysis in detail to provide a basis for subsequent decision-making.

[0033] In this embodiment, step S5 includes the following steps: Step S51: performing a drivable behavior space analysis according to the dynamic vehicle scenario twin model to generate a drivable behavior space for the current scenario; Step S52: making a dynamic emergency avoidance behavior decision on the scene dynamic emergency risk data according to the current scene drivable behavior space to generate emergency avoidance behavior data; Step S53: Perform safe driving path planning based on the scenario dynamic emergency risk data, thereby generating a safe driving path in the scenario; Step S54: Visualize the decision execution of the emergency avoidance behavior data and the safe driving path in the scene, so as to generate a visualization view of the emergency risk decision.

[0034] In this embodiment, the previously constructed dynamic vehicle scenario twin model is used to ensure that it contains information about the vehicle, surrounding environment, road structure and traffic participants. The model should have the ability to update in real time to reflect the current driving status and environmental changes, integrate sensor data (such as cameras, LiDAR and GPS) to enhance the visualization and analysis capabilities of the model, ensure the real-time and accuracy of the data, determine the definition of the drivable behavior space, including the vehicle's range of motion, road conditions, obstacle locations and the behavior of other traffic participants. The drivable behavior space should include areas where the vehicle can drive safely. Set analysis parameters such as safety distance, turning radius and acceleration limit to ensure that the generated drivable behavior space meets the actual driving conditions. Use space Analyze dynamic vehicle scenarios using analytical algorithms (such as Voronoi diagrams or Delaunay triangulations) to identify drivable and non-drivable areas. Determine safe driving paths through a comprehensive assessment of the environment. Use graphical algorithms (such as A* or Dijkstra) for path planning to ensure that the optimal path is found in the drivable behavior space. Generate a drivable behavior space map for the current scenario and identify safe driving areas and potential dangerous areas. Use visualization tools (such as Matplotlib or Open3D) to display analysis results to facilitate driving decision-making. Integrate the drivable behavior space of the current scenario with the scenario dynamic emergency risk data to ensure that the decision-making basis for each time step is Comprehensive, including information on potential obstacles, pedestrians, and traffic signals, prepare input data for the emergency avoidance decision model, including vehicle status, drivable behavior space, dynamic risk level, and surrounding traffic conditions, select appropriate decision models, such as rule-based systems, fuzzy logic control, or reinforcement learning models, to make decisions on emergency avoidance behaviors. Reinforcement learning models can optimize the decision-making process through training and improve the accuracy of emergency responses. Set the model's learning parameters and decision thresholds to ensure that the model can respond quickly to dynamic risks. Based on the integrated data, use the selected decision model to make dynamic emergency avoidance behavior decisions, calculate the priority of avoidance behaviors in real time, select the optimal avoidance plan (such as deceleration, lane change, or braking), and record each decision-making process. State information, including vehicle speed, acceleration, and decision timestamps, for subsequent analysis; Generate emergency avoidance behavior data reports; Record avoidance behavior decisions at each time step in detail, including decision basis and behavior selection; Integrate scene dynamic emergency risk data with vehicle state information to ensure that path planning can be based on comprehensive information, including the vehicle's current speed, direction, and dynamic changes in the surrounding environment; Prepare input data for the path planning model, including drivable behavior space, dynamic risk level, and obstacle location; Select an appropriate path planning model, such as A algorithm, RRT (rapid random tree), or D algorithm, to generate a safe driving path; A* algorithm is suitable for static environments, while RRT is suitable for dynamic environments; Set the parameters of the model,Such as heuristic functions and search ranges to ensure the effectiveness of path planning. Based on the integrated data, use the selected path planning model to generate a safe driving path. Calculate the shortest path and safe path in real time to ensure that the path avoids potential risk areas. Record the status information of each path planning process, including path length, risk assessment and timestamp, for subsequent analysis. Generate a visualization of the safe driving path, identify the starting point, end point and key nodes of the path, use visualization tools to display the results of path planning, and facilitate the evaluation of the safety and effectiveness of the path. Integrate emergency avoidance behavior data with safe driving path data to ensure that the visualization results can fully reflect the decision-making process, including the vehicle's driving status, avoidance behavior and path planning information. Prepare the input data of the visualization model, including vehicle trajectory, avoidance behavior and safe path. Select appropriate visualization tools, such as Matplotlib, Plotly or Unity 3D, to generate interactive visualization views. Unity 3D can provide more intuitive and dynamic visualization effects. You can set visualization parameters such as view angle, zoom ratio and animation effects to observe dynamic changes in the decision-making process. Based on the integrated data, use the selected visualization tool to generate a decision-making execution visualization view, display the vehicle's driving path, avoidance behavior and dynamic risk area together to ensure the comprehensiveness of the information. The visualization results are dynamically updated to reflect the vehicle's decision-making process and driving status in real time, enhancing interactivity.

[0035] In this embodiment, step S6 includes the following steps: Step S61: Identify the final decision action link of the emergency risk decision visualization view and extract each emergency action link; Step S62: performing instant perception feedback on the vehicle-mounted sensor based on each emergency action link, thereby generating an instant sensor feedback signal; Step S63: performing end-to-end decision optimization on the instant feedback signal of the sensor, thereby generating instant feedback end-to-end optimization data; Step S64: Perform global self-optimization of the vehicle on the instant feedback end-to-end optimization data to build a vehicle intelligent global decision-making engine.

[0036] In this embodiment, the emergency risk decision visualization view is loaded into the analysis platform to ensure that the view can dynamically display each link in the decision-making process, set the analysis target, and clearly identify the emergency action links that need to be identified, including avoidance, deceleration, parking and other behaviors, as well as the key moments of interaction with the environment. The visualization view is analyzed using image processing technology (such as OpenCV) to identify the key links in the decision-making process. The time and space information involved in different emergency actions are clarified through edge detection and contour extraction. The time series analysis method is used in combination with timestamp data to accurately identify the time and duration of each emergency action. The emergency actions in the visualization view are analyzed using rule-based algorithms or machine learning technologies (such as support vector machines or decision trees). Classify the actions, extract the characteristics of each link, set classification standards, such as action type, time range and environmental status, to ensure the accuracy and operability of the extracted action links, generate emergency action link identification reports, record the start time, end time and related feature information of each action link in detail, integrate the extracted emergency action links with the data streams of on-board sensors, including data from sensors such as cameras, radars and LiDAR, to ensure that the real-time data of the sensors can match the emergency action links, so as to obtain the necessary environmental information when performing emergency actions, and configure the working mode of the sensor according to the characteristics of the emergency action links. For example, when decelerating, increase the field of view of the camera to capture more details of the road surface to ensure safety. Set the sampling frequency and trigger conditions of the sensor to obtain instant feedback at critical moments, use data fusion algorithms (such as Kalman filter) to process sensor data and generate instant feedback signals, ensure the accuracy and reliability of feedback signals by fusing data from different sensors, set the format of feedback signals, including signal type (such as speed, acceleration, distance, etc.) and timestamp, ensure the traceability of signals, generate sensor instant feedback signal reports, record the generation time, type and characteristics of each feedback signal, use visualization tools to display the changing trend of feedback signals, so as to facilitate analysis and optimization of subsequent decision-making processes, integrate sensor instant feedback signals with current vehicle status information (such as speed, direction and acceleration) to ensure optimization of models Make decisions based on comprehensive data, prepare input data for the optimization model, including sensor feedback signals, emergency action links, and vehicle status information, select appropriate decision optimization models, such as reinforcement learning (RL) algorithm or deep Q network (DQN), to achieve end-to-end decision optimization. Reinforcement learning can self-adjust strategies through real-time feedback to improve decision accuracy. Set the learning parameters of the optimization model, such as learning rate, discount factor, and number of training rounds, to ensure the stability and effectiveness of the model. Use historical data to train the selected optimization model and verify the effect of the model on the test set. Ensure effective decision optimization in different scenarios by adjusting model parameters. Perform real-time training and use instant feedback signals from sensors to optimize the model online.Improve the adaptability of the model, input the integrated data into the trained optimization model, perform end-to-end optimization with instant feedback, record the optimization decision results of each time step, including the optimized vehicle status and actions, generate an instant feedback end-to-end optimization data report, record the key parameters and decision results in the optimization process in detail, provide a basis for subsequent analysis, integrate the instant feedback end-to-end optimization data with the historical decision data, ensure that the global decision engine makes decisions based on comprehensive data, prepare the input data of the engine, including vehicle status, environmental information, current risk assessment and instant feedback data, design the architecture of the intelligent global decision engine, including data input module, decision optimization module and output execution module, ensure that the engine can respond to and handle complex traffic situations in real time, set the engine's working mode The engine should be able to adjust the decision-making strategy according to the real-time environment changes, improve the accuracy and security of the decision-making, record the key parameters and decision results during the operation of the engine, generate a global self-optimization report, and provide a basis for subsequent system iterations.

[0037] In this embodiment, an intelligent decision support system for vehicle-mounted sensor data fusion is provided, which is used to execute the intelligent decision support method for vehicle-mounted sensor data fusion as described above, including: The heterogeneous fusion module is used to obtain multi-source vehicle operation status monitoring data based on vehicle-mounted sensors; the vehicle state heterogeneous fusion and multi-period change analysis of the vehicle operation status monitoring data are performed to obtain the evolution trend characteristics of the vehicle operation status; The running scene module is used to obtain the vehicle driving recorder monitoring image and vehicle dynamic environment monitoring parameters, and perform multi-target visual detection and dynamic running scene fitting to construct the vehicle running scene diagram; The twin model module is used to perform three-dimensional scene point cloud modeling on the vehicle operation scene graph, and perform dynamic vehicle trend rendering according to the evolution trend characteristics of the vehicle operation state, thereby obtaining a dynamic vehicle scene twin model; The operation simulation module is used to perform dynamic vehicle operation simulation on the dynamic vehicle scenario twin model, and then perform dynamic emergency risk analysis to generate scenario dynamic emergency risk data; The emergency decision-making module is used to make dynamic emergency avoidance behavior decisions and safe driving path planning based on the scene dynamic emergency risk data, thereby generating a visual view of the emergency risk decision; The instant feedback optimization module is used to provide instant perception feedback based on the emergency risk decision visualization view, perform global self-optimization of the vehicle, and build an intelligent global decision-making engine for the vehicle.

[0038] The present invention can obtain data of different dimensions from different types of vehicle-mounted sensors (such as radar, camera, GPS, acceleration sensor, etc.) and fuse them through the heterogeneous fusion module. This means that the system can take advantage of the data of various sensors to form a full range of more comprehensive vehicle status information. By analyzing the operation data of multiple time periods, the module identifies the change trend of vehicle status over time (such as speed change, driving habits, driving environment, etc.). The extraction of this trend feature helps to predict future vehicle status changes and provide data support for subsequent decision-making. Through the combination of driving recorders and dynamic environmental monitoring parameters, the system can perceive various factors on the road in real time (such as pedestrians, etc.). , other vehicles, road obstacles, etc.), so as to fully grasp the dynamic environment around the vehicle. Through multi-target visual detection, the module can identify multiple potential dangerous targets in dynamic scenes (such as other vehicles, pedestrians, traffic signs, etc.), and combine this information with the current driving scene. Such multi-target perception capability greatly enhances the real-time response capability of the system. The 3D scene point cloud modeling more realistically restores the environment in which the vehicle is located, including elements such as roads, surrounding vehicles, obstacles, etc. Through high-precision 3D modeling, the vehicle better understands and predicts changes in the surrounding environment. According to the evolution trend characteristics of the vehicle's operating state, the twin model module can render dynamic trends in real time and simulate different driving states. Through this rendering, the system predicts the driving path and potential dangers in advance and provides decision support. By running simulations, the system can simulate different driving scenarios, conduct virtual experiments, and identify potential traffic risks in advance. This process helps to evaluate the risks of different decision-making plans and ensure the selection of the best decision. During the simulation process, the system can identify and analyze the probability of various emergency situations (such as sudden braking of vehicles, collisions, etc.). Through emergency risk analysis, the system formulates appropriate emergency response strategies to improve driving safety. Based on real-time risk data, the system can make rapid emergency decisions (for example: automatic braking, obstacle avoidance, lane changes, etc.) to ensure that the vehicle can respond quickly in dangerous situations and reduce Accident risk, in addition to emergency avoidance decisions, the system can also plan safe routes to avoid risk areas such as congested areas, traffic accidents, bad weather, etc., to improve driving safety and smoothness. The vehicle's intelligent global decision-making engine can comprehensively consider all perception, simulation, and decision-making data to globally optimize the vehicle's behavior. This optimization is not limited to individual scenarios, but covers various situations in the entire driving process, ensuring that the system can adaptively handle complex and changing road environments. Through global optimization, the vehicle can intelligently adjust its driving strategy to avoid traffic accidents, save driving time, and improve traffic smoothness. For example, when congestion or an accident occurs, the system can intelligently adjust the driving route to avoid blocked areas.

[0039] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0040] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. An intelligent decision support method for vehicle-mounted sensor data fusion, characterized in that: The following steps are involved: Step S1: Acquire vehicle multi-source operating status monitoring data based on vehicle-mounted sensors; Carry out heterogeneous fusion of vehicle status and multi-period change analysis of vehicle status monitoring data from multiple sources to obtain the evolution trend characteristics of vehicle status; Step S2: Acquire the vehicle driving recorder monitoring image and vehicle dynamic environment monitoring parameters, perform multi-target visual detection and dynamic operation scene fitting, and construct a vehicle operation scene diagram; Step S3: Modeling the three-dimensional scene point cloud of the vehicle running state scene graph, and rendering the dynamic vehicle trend according to the evolution trend characteristics of the vehicle running state, so as to obtain a dynamic vehicle scene twin model; Step S4: Perform dynamic vehicle operation simulation on the dynamic vehicle scenario twin model, and then perform dynamic emergency risk analysis to generate scenario dynamic emergency risk data; Step S5: Perform dynamic emergency avoidance behavior decision and safe driving path planning on the scene dynamic emergency risk data, thereby generating a visual view of the emergency risk decision; Step S6: Based on the emergency risk decision visualization view, instant perception feedback is performed, and vehicle global self-optimization is performed to build a vehicle intelligent global decision engine.

2. The intelligent decision support method for vehicle-mounted sensor data fusion according to claim 1 is characterized in that: The specific steps of step S1 are: Step S11: acquiring vehicle multi-source operating status monitoring data based on vehicle-mounted sensors; Step S12: performing vehicle state heterogeneous fusion on vehicle multi-source operation state monitoring data, thereby generating multi-dimensional vehicle operation state features; Step S13: performing a multi-period change analysis on the multi-dimensional running state characteristics of the vehicle to generate the running state characteristics of the vehicle in multiple time periods; Step S14: Evolving the state change trend of the vehicle running state characteristics in multiple time periods, thereby obtaining the vehicle running state evolution trend characteristics.

3. The intelligent decision support method for vehicle-mounted sensor data fusion according to claim 2 is characterized in that: The specific steps of step S12 are: The vehicle multi-source operating status monitoring data includes vehicle dynamics data, engine monitoring data, battery status monitoring data, tire pressure sensor parameters and IMU real-time monitoring parameters; Perform vehicle speed change analysis on vehicle dynamics data to obtain vehicle time-series speed change characteristics; Perform engine temperature change identification on engine monitoring data and extract engine temperature change data; Perform discrete temperature variation fitting on the engine temperature variation data to construct the engine temperature variation curve; Perform real-time load analysis on battery status monitoring data to obtain real-time battery load status data; Perform tire pressure stability identification on tire pressure sensing parameters to generate vehicle tire pressure operation balance characteristics; Carry out body posture fluctuation detection on the IMU real-time monitoring parameters to generate body posture fluctuation data; The vehicle state heterogeneous fusion is performed on the vehicle's time-series speed change characteristics, engine temperature change curve, battery real-time load status data and body posture fluctuation data to generate multi-dimensional vehicle operation state characteristics.

4. The intelligent decision support method for vehicle-mounted sensor data fusion according to claim 1 is characterized in that: The specific steps of step S2 are: Step S21: Acquire the vehicle driving recorder monitoring image and vehicle dynamic environment monitoring parameters; Step S22: Analyzing the current environment change characteristics based on the vehicle dynamic environment monitoring parameters to obtain the current vehicle operating environment change characteristics; Step S23: performing multi-target visual detection on the vehicle driving recorder monitoring image and marking each scene target; Step S24: performing deep semantic analysis on each scene object, thereby generating semantic features of each scene object; Step S25: performing real-time operation scenario analysis based on the semantic features of each scenario target to generate vehicle real-time operation scenario data; Step S26: Perform dynamic operation scenario fitting on the real-time vehicle operation scenario data based on the current vehicle operation environment change characteristics to construct a vehicle operation state scenario diagram.

5. The intelligent decision support method for vehicle-mounted sensor data fusion according to claim 1 is characterized in that: The specific steps of step S3 are: Step S31: accurately calculating the spatial position of each scene object to obtain the spatial position coordinates of each scene object; Step S32: performing a three-dimensional road structure analysis on the vehicle running state scene graph, thereby generating three-dimensional structure data of the running scene; Step S33: performing three-dimensional scene point cloud modeling on the three-dimensional structure data of the running scene based on the spatial position coordinates of each scene object, and constructing a real-time three-dimensional scene model of the vehicle; Step S34: Perform dynamic vehicle trend rendering on the real-time three-dimensional scene model of the vehicle according to the evolution trend characteristics of the vehicle operation state, so as to obtain a dynamic vehicle scene twin model.

6. The intelligent decision support method for vehicle-mounted sensor data fusion according to claim 1 is characterized in that: The specific steps of step S4 are: Step S41: performing dynamic vehicle operation simulation on the dynamic vehicle scenario twin model to generate vehicle operation simulation response data; Step S42: pre-identify the road condition of the dynamic vehicle scene twin model and extract the road condition data ahead; Step S43: predicting the current vehicle driving situation based on the vehicle operation simulation response data based on the front road condition data to generate vehicle driving situation prediction data; Step S44: performing road vehicle operation trend prediction on the vehicle operation simulation response data to obtain road vehicle trend prediction data; Step S45: calculating the probability of pedestrian crossing the road based on the vehicle operation simulation response data, thereby generating the probability of pedestrian crossing the road; Step S46: Perform dynamic emergency risk analysis based on the road vehicle trend prediction data, the road pedestrian crossing probability and the vehicle driving situation prediction data, so as to generate scene dynamic emergency risk data.

7. The intelligent decision support method for vehicle-mounted sensor data fusion according to claim 1 is characterized in that: The specific steps of step S5 are: Step S51: performing a drivable behavior space analysis according to the dynamic vehicle scenario twin model to generate a drivable behavior space for the current scenario; Step S52: making a dynamic emergency avoidance behavior decision on the scene dynamic emergency risk data according to the current scene drivable behavior space to generate emergency avoidance behavior data; Step S53: Perform safe driving path planning based on the scenario dynamic emergency risk data, thereby generating a safe driving path in the scenario; Step S54: Visualize the decision execution of the emergency avoidance behavior data and the safe driving path in the scene, so as to generate a visualization view of the emergency risk decision.

8. The intelligent decision support method for vehicle-mounted sensor data fusion according to claim 1 is characterized in that: The specific steps of step S6 are: Step S61: Identify the final decision action link of the emergency risk decision visualization view and extract each emergency action link; Step S62: performing instant perception feedback on the vehicle-mounted sensor based on each emergency action link, thereby generating an instant sensor feedback signal; Step S63: performing end-to-end decision optimization on the instant feedback signal of the sensor, thereby generating instant feedback end-to-end optimization data; Step S64: Perform global self-optimization of the vehicle on the instant feedback end-to-end optimization data to build a vehicle intelligent global decision-making engine.

9. An intelligent decision support system for vehicle-mounted sensor data fusion, characterized in that: The intelligent decision support method for performing the vehicle-mounted sensor data fusion as claimed in claim 1 comprises: The heterogeneous fusion module is used to obtain multi-source vehicle operation status monitoring data based on vehicle-mounted sensors; the vehicle state heterogeneous fusion and multi-period change analysis of the vehicle operation status monitoring data are performed to obtain the evolution trend characteristics of the vehicle operation status; The running scene module is used to obtain the vehicle driving recorder monitoring image and vehicle dynamic environment monitoring parameters, and perform multi-target visual detection and dynamic running scene fitting to construct the vehicle running scene diagram; The twin model module is used to perform three-dimensional scene point cloud modeling on the vehicle operation scene graph, and perform dynamic vehicle trend rendering according to the evolution trend characteristics of the vehicle operation state, thereby obtaining a dynamic vehicle scene twin model; The operation simulation module is used to perform dynamic vehicle operation simulation on the dynamic vehicle scenario twin model, and then perform dynamic emergency risk analysis to generate scenario dynamic emergency risk data; The emergency decision-making module is used to make dynamic emergency avoidance behavior decisions and safe driving path planning based on the scene dynamic emergency risk data, thereby generating a visual view of the emergency risk decision; The instant feedback optimization module is used to provide instant perception feedback based on the emergency risk decision visualization view, perform global self-optimization of the vehicle, and build an intelligent global decision-making engine for the vehicle.

Images