Data visualization method and system for intelligent automobile data recorder of electric vehicle

Abstract

The invention relates to the field of data visualization, in particular to a data visualization method and system for an intelligent automobile data recorder of an electric vehicle. The method comprises the following steps: acquiring a real-time video stream of a driving scene based on an intelligent automobile data recorder, performing motion visual jitter elimination compensation and multi-level scene target feature perception, and constructing a multi-level scene target information sequence; performing space-time connection rendering according to the multi-level scene target information sequence, and performing scene evolution smoothing processing so as to construct a dynamic driving scene topology rendering model; vehicle running state parameters are obtained, multi-parameter correlation analysis and vehicle transient response analysis are carried out, and a real-time vehicle state evaluation report is constructed; and performing multi-target constraint calculation based on the dynamic driving scene topology rendering model, and performing optimal vehicle speed prediction to obtain the optimal vehicle speed. According to the invention, through multi-dimensional driving visual rendering, good real-time feedback and behavior early warning are provided for the user.

Description

Technical Field

[0001] The present invention relates to the field of data visualization, and in particular to a data visualization method and system for an intelligent driving recorder of an electric vehicle. Background Art

[0002] With the continuous development of intelligent transportation, the Internet of Things, and new energy technologies, electric vehicles, as green and environmentally friendly means of transportation, have been widely used in various scenarios, including urban transportation, shared mobility, and logistics distribution. At the same time, the demand for data collection, analysis, and management during the operation of electric vehicles is also increasing. This is particularly true in areas such as vehicle safety monitoring, driving behavior analysis, and operational scheduling optimization, which place higher demands on high-precision, real-time data recording and visual analysis. As a key terminal device for data collection, smart driving recorders have gradually become standard equipment for electric vehicles. Their functions are not limited to traditional video recording and accident evidence collection, but have extended to multi-dimensional intelligent functions such as trajectory tracking, driving behavior recognition, energy consumption monitoring, and communication linkage.

[0003] Electric vehicles generate a vast amount of heterogeneous data during operation, including video images, geolocation information, speed fluctuations, battery status, and driving behavior parameters. Without effective integration and visualization, this data will struggle to realize its true value in ensuring driving safety and supporting operational decision-making. Traditional data display methods and analysis tools are particularly inadequate for the increasingly complex operational management requirements of electric vehicles, particularly when faced with high-frequency, high-dimensional, and multi-source data streams. Therefore, developing an efficient, intuitive, and intelligent data visualization method has become crucial for enhancing the value of smart dashcams.

[0004] Currently, existing smart dashcams for electric vehicles often use static charts, tabular information, or simple graphical interfaces to present data. While these methods offer some information transmission capabilities, they often lack interactivity and dynamic responsiveness, making it difficult to reflect changing trends and potential risks in real time. Furthermore, these visualization methods are often developed in isolation, lacking the ability to deeply integrate multi-source data and identify behavioral patterns. This makes it difficult to support advanced intelligent functions such as driving behavior analysis, fault prediction and warning, and operational optimization decisions in practical applications. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention proposes a data visualization method and system for an electric vehicle intelligent driving recorder to solve at least one of the above technical problems.

[0006] To achieve the above object, the present invention provides a data visualization method for an electric vehicle intelligent driving recorder, comprising the following steps: Step S1: Acquire a real-time video stream of the driving scene based on the intelligent driving recorder, perform motion visual jitter elimination compensation and multi-level scene target feature perception, and construct a multi-level scene target information sequence; Step S2: Perform spatiotemporal connection rendering based on the multi-level scene target information sequence and perform scene evolution smoothing to construct a dynamic driving scene topology rendering model; Step S3: Obtain vehicle operating status parameters, perform multi-parameter correlation analysis and vehicle transient response analysis, and construct a real-time vehicle status assessment report; Step S4: performing multi-objective constraint calculation based on the dynamic driving scene topology rendering model, and performing optimal vehicle speed prediction based on the real-time vehicle status assessment report to obtain the optimal vehicle speed; Step S5: Calculate the multi-path probability of dynamic targets on the dynamic driving scene topology rendering model, perform credibility weight analysis, and extract the maximum probability behavior prediction trajectory of each target; Step S6: Calculate the optimal safety distance for each target based on the maximum probability behavior prediction trajectory and the optimal vehicle speed, and perform visualization rendering on the vehicle's central control screen to perform visualization operations.

[0007] In this specification, a data visualization system for an electric vehicle intelligent driving recorder is provided, which is used to execute the data visualization method for the electric vehicle intelligent driving recorder as described above, including: The video processing module is used to obtain real-time video streams of driving scenes based on intelligent driving recorders, perform motion visual jitter elimination and compensation, and perceive multi-level scene target features to construct a multi-level scene target information sequence; The scene topology rendering module is used to perform spatiotemporal connection rendering based on multi-level scene target information sequences and perform scene evolution smoothing to construct a dynamic driving scene topology rendering model; The vehicle status assessment module is used to obtain vehicle operating status parameters, perform multi-parameter correlation analysis and vehicle transient response analysis, and construct a real-time vehicle status assessment report; A multi-objective constraint module is used to perform multi-objective constraint calculations based on a dynamic driving scenario topology rendering model and to predict the optimal vehicle speed based on a real-time vehicle status assessment report to obtain the optimal vehicle speed. The behavior trajectory prediction module is used to calculate the multi-path probability of dynamic targets in the dynamic driving scene topology rendering model, perform credibility weight analysis, and extract the maximum probability behavior prediction trajectory for each target; The visualization rendering module is used to calculate the optimal safety distance for each target based on the maximum probability behavior prediction trajectory and the optimal vehicle speed, and perform visualization rendering on the vehicle's central control screen to perform visualization operations.

[0008] The beneficial effects of the present invention include: A high-frame-rate, high-definition, real-time driving scene video stream is captured through an intelligent driving recorder, ensuring excellent visual quality of the data source. To address image jitter caused by uneven roads, sharp turns, and vibrations during electric vehicle driving, motion visual jitter removal and image stabilization algorithms (such as those based on optical flow or deep learning) are introduced to effectively improve the continuity and visual stability of the video stream. Furthermore, multi-level object detection and feature perception are performed to semantically segment and annotate key objects in the driving scene, such as pedestrians, vehicles, lane markings, traffic signs, non-motorized vehicles, and obstacles, and extract object location, size, type, motion direction, and speed information. By organizing this object information in a hierarchical manner, a multi-level scene object information sequence is constructed, providing a structured, high-semantic density input data source for subsequent scene modeling, behavior deduction, and visualization. By constructing dynamic connections between object states along a timeline, the continuous scene evolution process is restored and the underlying interactive behaviors and evolutionary trends within the driving scene are revealed. At the same time, spatiotemporal interpolation and evolutionary smoothing algorithms, such as image displacement field interpolation and multi-frame progressive synthesis, are introduced to effectively address scene discontinuities caused by sudden changes in the driving environment and objects entering or leaving the field of view, thereby improving the coherence of scene transitions and the visual naturalness of rendering. By accessing onboard system interfaces (such as OBD or CAN bus), key operating parameters, including vehicle speed, battery voltage, motor power output, steering angle, acceleration and deceleration, and braking signals, are collected in real time to comprehensively monitor the vehicle's current operating status. Multi-parameter cross-analysis techniques (such as covariance analysis and cluster regression) are then used to reveal the response relationships between these operating parameters. Transient response modeling is then used to analyze short-term vehicle dynamic behavior, such as transient changes in sudden acceleration, emergency braking, and offset steering. Multi-objective constraint computational models (such as those based on Markov decision processes or reinforcement learning frameworks) are used to identify risk points, dynamic obstacles, and drivable areas within the current driving scene. Combined with the current vehicle state, the controllability and risk level at different speeds are comprehensively assessed to derive the optimal vehicle speed prediction that meets the current driving environment and vehicle capabilities. This speed not only meets safe driving requirements but also optimizes energy consumption and improves travel efficiency, forming a key foundation for intelligent assisted driving control and human-machine collaborative operation. By performing trajectory prediction analysis on each moving target (such as the vehicle ahead, crossing pedestrians, and electric vehicles) in a dynamic driving scenario, refined target behavior prediction is achieved. Based on a dynamic scene topology rendering model, combining historical behavior trajectories with current dynamic parameters (speed, direction, acceleration, etc.), a multi-path behavior prediction algorithm (such as those based on Bayesian networks and multi-modal trajectory prediction transformers) is employed to deduce multiple possible target motion paths in the short term.A credibility weighting analysis mechanism is further introduced, assigning a different credibility score to each predicted path based on factors such as target behavior stability, scenario complexity, and mutual interference. Ultimately, the target trajectory with the highest predicted probability in the current scenario is extracted, serving as the basis for the system's proactive risk avoidance, path adjustment, and speed control decisions. A real-time safety distance tensor matrix is ​​dynamically generated based on factors such as target type (e.g., pedestrian, vehicle), relative speed, and directional angle. Combined with risk assessments at different levels, the system dynamically adjusts the safety distance and provides spatial buffering for forward targets, lateral interference, and potential intersecting targets. Finally, a multi-dimensional driving visualization is rendered on the vehicle's central control screen using layered overlays, progressive visual guidance, and color coding. This visualization interface not only provides real-time feedback and behavioral warning capabilities, but also implements an intelligent transformation path from data to cognition to guidance, significantly enhancing the driver's perception and response efficiency in complex traffic environments, thereby improving overall driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 This is a schematic flow chart of the steps of a data visualization method for an electric vehicle intelligent driving recorder according to the present invention; Figure 2 Detailed implementation flow chart of step S1; Figure 3 Detailed implementation flow chart of step S2; Figure 4 Schematic diagram of the detailed implementation steps of step S3. DETAILED DESCRIPTION

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

[0011] This application provides a data visualization method and system for an intelligent driving recorder for an electric vehicle. The execution entities of the data visualization method and system for an intelligent driving recorder for an electric vehicle include, but are not limited to, the following: mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc., which can be considered as general computing nodes of this application. 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 4 The present invention provides a data visualization method for an electric vehicle intelligent driving recorder, the data visualization method for an electric vehicle intelligent driving recorder comprising the following steps: Step S1: Acquire a real-time video stream of the driving scene based on the intelligent driving recorder, perform motion visual jitter elimination compensation and multi-level scene target feature perception, and construct a multi-level scene target information sequence; Step S2: Perform spatiotemporal connection rendering based on the multi-level scene target information sequence and perform scene evolution smoothing to construct a dynamic driving scene topology rendering model; Step S3: Obtain vehicle operating status parameters, perform multi-parameter correlation analysis and vehicle transient response analysis, and construct a real-time vehicle status assessment report; Step S4: performing multi-objective constraint calculation based on the dynamic driving scene topology rendering model, and performing optimal vehicle speed prediction based on the real-time vehicle status assessment report to obtain the optimal vehicle speed; Step S5: Calculate the multi-path probability of dynamic targets on the dynamic driving scene topology rendering model, perform credibility weight analysis, and extract the maximum probability behavior prediction trajectory of each target; Step S6: Calculate the optimal safety distance for each target based on the maximum probability behavior prediction trajectory and the optimal vehicle speed, and perform visualization rendering on the vehicle's central control screen to perform visualization operations.

[0013] In the embodiment of the present invention, see Figure 1 , is a flowchart of the steps of a data visualization method for an electric vehicle intelligent driving recorder according to the present invention. In this example, the steps of the data visualization method for an electric vehicle intelligent driving recorder include: Step S1: Acquire a real-time video stream of the driving scene based on the intelligent driving recorder, perform motion visual jitter elimination compensation and multi-level scene target feature perception, and construct a multi-level scene target information sequence; In this embodiment, the smart dashcam is equipped with a high-resolution camera, typically 1080p or higher, to ensure clear video streams in all lighting conditions, such as daytime and nighttime. The video frame rate is set to 30 frames per second (FPS) to ensure smooth and continuous video. While the vehicle is in motion, the dashcam records the driving environment in real time, generating continuous video data. When capturing video, the dashcam should be securely mounted to avoid image blur or jitter caused by vehicle motion. Proper mounting and anti-vibration design ensure the stability of the video stream, providing a reliable data foundation for subsequent processing. After acquiring the real-time video stream, motion visual jitter compensation is performed to eliminate jitter caused by vehicle motion, ensuring a clear and stable output video stream. Image processing techniques can analyze changes between consecutive frames and identify jitter caused by vehicle motion. By comparing consecutive video frames, feature points are identified and their motion trajectories are calculated. Using optical flow or feature point tracking algorithms, the system tracks the changes of each feature point across consecutive frames to determine the overall jitter level. Once jitter signatures are identified, the system compensates for the video frames. Compensation methods typically include translation and rotation correction, adjusting each frame to an ideal position to eliminate jitter. This process not only improves video quality but also provides more accurate image data for subsequent object detection and analysis. After jitter correction, multi-layer scene object feature perception is performed. Information about various dynamic and static objects in the video stream is extracted to build a comprehensive sequence of scene object information. Advanced object detection algorithms, such as YOLO or Faster R-CNN, are applied to each frame to identify vehicles, pedestrians, traffic signs, and other objects in the scene. This process involves not only object category recognition but also localization and confidence assessment. The system stores and categorizes detected object information (such as category, location, and confidence). In a given frame, the system might detect a car, a pedestrian, and a traffic sign. Information about each object is recorded, including its specific location in the image and the confidence level of the recognition. This multi-layered object feature perception ensures that every important element of the driving scene is accurately captured, providing critical data for driving behavior analysis and safety assessment.

[0014] Step S2: Perform spatiotemporal connection rendering based on the multi-level scene target information sequence and perform scene evolution smoothing to construct a dynamic driving scene topology rendering model; In this embodiment, the objects in each frame are arranged in a temporal sequence using previously acquired multi-layered scene object information. This requires identifying the position and state of each object at different time points and integrating this information into a continuous sequence. The system must be able to process real-time data streams and associate each frame's video data with the corresponding object information to ensure accurate representation of the object's dynamic changes. Interpolation algorithms (such as linear interpolation or spline interpolation) are used to smoothly connect the object positions at different time points. This process eliminates sudden changes caused by sensor noise or image processing delays, making the object movement in dynamic scenes more natural. For example, if a vehicle moves from (x1, y1) to (x2, y2) in two consecutive frames, an interpolation algorithm can be used to calculate the intermediate position between these two time points, achieving a smooth transition. Furthermore, the system must account for background changes in the scene, such as road conditions and lighting variations. This can be achieved by performing layered background processing. By analyzing static elements in the scene (such as roads and buildings), the background is separated from dynamic objects to ensure a layered and realistic rendering effect. After completing the spatiotemporal connection rendering, scene evolution smoothing is performed. Rendering effects are further enhanced to ensure the smoothness and visual realism of dynamic driving scenes. Smoothing algorithms such as Kalman filters are applied to the target's trajectory for trajectory correction. Kalman filters effectively filter noise, thereby improving the accuracy of target position estimation. By predicting and updating the target's motion state at each moment, the system can obtain a more stable and accurate target path. Next, the target's kinematic parameters (such as velocity and acceleration) are combined to analyze and predict its behavior. This process uses a physical model to simulate the target's motion characteristics, achieving a more natural behavior. The vehicle's trajectory during acceleration and deceleration can be described using quadratic equations, making the rendering more realistic. This process also requires consideration of the interaction between the target and the environment, such as the vehicle's roll effect when turning and the dynamic reaction of pedestrians crossing a zebra crossing. These factors can affect the scene's realism and must be accounted for in the rendering model. The results of these steps are integrated to construct a topological rendering model for the dynamic driving scene. This generates a complete and interactive dynamic scene, facilitating subsequent data visualization and analysis. A topological rendering model for dynamic driving scenarios should possess the following characteristics: Real-time: The model can update scene objects and environmental information based on real-time data, ensuring that the rendered content matches the actual driving situation. Interactivity: Users can interact with the scene through an interface, such as zooming in, out, and rotating the view. Clear hierarchy: Dynamic objects and static backgrounds are clearly distinguished, making key information in complex scenes easy to identify. During the construction process, 3D modeling software such as Unity or Unreal Engine can be used to build and render the scene. These software programs provide a wealth of tools and libraries that can efficiently handle complex dynamic scenes.The successfully constructed dynamic driving scene topology rendering model will provide visualization support for the intelligent management and safe driving of electric vehicles, enabling drivers to obtain key information in real time and improve the overall driving experience.

[0015] Step S3: Obtain vehicle operating status parameters, perform multi-parameter correlation analysis and vehicle transient response analysis, and construct a real-time vehicle status assessment report; In this embodiment, during data acquisition, an OBD-II adapter is connected to the vehicle's OBD port to read real-time vehicle data, including vehicle speed, engine speed, throttle position, brake status, coolant temperature, and more. The Python python-OBD library is used to communicate with the OBD-II adapter and obtain real-time vehicle status parameters. In the specific implementation, the OBD connection is first initialized to ensure smooth communication with the vehicle. Then, various vehicle status parameters are obtained by sending requests. After each successful read, the data is stored in a dictionary or data frame for subsequent analysis. During the process of obtaining vehicle status parameters, the connection status needs to be monitored to handle possible failures or data loss. The real-time and accuracy of the data is ensured. The resulting vehicle operating status parameters will provide basic data for subsequent multi-parameter correlation analysis. Correlations between different operating parameters are identified to discover potential patterns and anomalies. Experimental parameters are set, such as a 95% confidence level for the correlation analysis. Data analysis is performed using the Python Pandas and SciPy libraries. The obtained vehicle status parameters are organized into a data frame format, ensuring that data for each parameter within the same time period is aligned. The pandas.DataFrame.corr() function is used to calculate the correlation coefficient matrix between each parameter, obtaining the linear correlation between the parameters. During the analysis, particular attention is paid to parameter combinations with significant correlations, and these correlation patterns are recorded. The correlation between throttle opening and vehicle speed may indicate the vehicle's acceleration performance, while the correlation between engine speed and power output may reflect the engine's operating status. Using visualization tools such as Seaborn, heat maps are created to visually display the correlations between different parameters. This process not only helps identify normal operating modes but also detects anomalous data points, providing important evidence for subsequent vehicle transient response analysis. After completing the multi-parameter correlation analysis, vehicle transient response analysis is conducted to evaluate the vehicle's dynamic response performance under different driving behaviors. By analyzing changes in vehicle operating conditions, the transient response characteristics of the vehicle are understood. Experimental parameters are set, such as a 5-second time window for the response analysis, to capture transient changes. When conducting transient response analysis, the current driving behavior, such as sudden acceleration, sudden braking, or constant speed driving, is first identified. Combined with previously acquired vehicle state parameters, each parameter is smoothed using time series analysis methods (such as moving average or exponential smoothing) to extract key response characteristics (such as response time and peak change). In practice, the data can be numerically integrated using the NumPy and SciPy libraries to calculate acceleration and velocity changes within each time period. This allows for evaluating the vehicle's responsiveness under different driving behaviors. In the case of sudden braking, the relationship between brake pedal pressure and vehicle speed reduction can be analyzed to assess braking effectiveness.The resulting vehicle transient response data will provide crucial information for subsequent real-time vehicle status assessment reports. By combining transient response characteristics with multi-parameter correlation analysis results, a more comprehensive understanding of the vehicle's dynamic performance can be achieved. After completing the vehicle transient response analysis, integrate the previous analysis results to construct a real-time vehicle status assessment report. Organize all analysis results into a structured report to facilitate understanding of the vehicle's operating status for drivers and managers. Set experimental parameters, such as setting the report update frequency to once per minute, to ensure data timeliness. Summarize the acquired vehicle status parameters, multi-parameter correlation analysis results, and transient response analysis data and organize them into a data frame using a Python data processing library (such as Pandas). The report should include the current value of each parameter, correlation analysis results, and transient response characteristics. Convert the organized data frame into a visualization format. Use a visualization library (such as Matplotlib or Plotly) to generate charts that display vehicle status trends and key performance indicators. Use a line chart to display changes in vehicle speed and rotational speed, and a bar chart to illustrate the correlations between parameters. The generated real-time vehicle status assessment report is exported in PDF or HTML format, ensuring easy sharing and viewing. This report provides drivers with real-time vehicle status information, helping them evaluate vehicle performance and make appropriate decisions. This series of steps ensures the effective implementation of data visualization methods for electric vehicle smart dashcams, providing a scientific basis for safe driving.

[0016] Step S4: performing multi-objective constraint calculation based on the dynamic driving scene topology rendering model, and performing optimal vehicle speed prediction based on the real-time vehicle status assessment report to obtain the optimal vehicle speed; In this embodiment, multi-objective constraint calculations are performed based on a dynamic driving scenario topology rendering model. By analyzing the environment and vehicle state, multiple constraints are calculated for the current driving situation, providing a basis for subsequent optimal vehicle speed prediction. Experimental parameters are set, such as a 5-second time window for constraint calculations, to ensure data timeliness. When implementing multi-objective constraint calculations, the various constraints affecting vehicle travel must first be identified and defined. These factors include, but are not limited to, road speed limits, traffic density, the behavior of surrounding dynamic targets, road complexity, and weather conditions. Speed ​​limit information can be obtained through the previous speed limit sign recognition step, and traffic density can be determined through real-time dynamic target detection. Next, the overall constraint conditions are calculated using data from the vehicle state assessment report and the aforementioned constraints. A weighting strategy can be used to assign a weight to each constraint factor to reflect its importance in speed prediction. Speed ​​limit information may have a higher weight, while weather conditions may have a relatively lower weight. During the specific calculation, a linear programming method can be used to transform each constraint into a mathematical model. The objective function is set to the optimal speed and the solution is obtained within the constraints. The optimization module in Python's SciPy library can be used to solve the problem by calling the linprog function. The resulting multi-objective constraint calculation results provide the necessary data support for subsequent optimal speed prediction, ensuring safe and efficient vehicle operation. After the multi-objective constraint calculation is completed, the optimal vehicle speed prediction is performed based on the real-time vehicle state assessment report. Leveraging previously acquired vehicle state and constraint information, the optimal speed is intelligently predicted for the current driving environment. Experimental parameters are set, such as using a support vector machine (SVM) as the speed prediction model, to improve prediction accuracy. When preparing the speed prediction, the data from the real-time vehicle state assessment report must be formatted for machine learning model input. This data should include key parameters such as the vehicle's current speed, acceleration, throttle position, and brake status, as well as the constraints obtained from the multi-objective constraint calculation. A training set is constructed to train the speed prediction model. Historical driving data can be used, including speed records and corresponding vehicle state parameters under various driving scenarios. This data is divided into training and test sets using the train_test_split function in the scikit-learn library. Once the data is prepared, the support vector machine model is trained. Select an appropriate kernel function (such as an RBF kernel) and set appropriate hyperparameters, then use the SVC class to train the model. After training, evaluate the model's performance using a test set to ensure prediction accuracy meets the established standards (root mean square error (RMSE) less than 5 km / h). After model training, input the real-time vehicle state parameters and calculated constraints into the support vector machine model to predict the optimal speed. The goal of this process is to provide a safe and efficient speed recommendation based on the current driving environment and vehicle state.In implementation, the predict method can be used to predict the input data and generate the optimal vehicle speed. Based on the model's output, a reasonable range (e.g., ±5 km / h) is set for adjustment to ensure the recommended speed is within a safe range. Furthermore, the predicted speed can be further adjusted based on the results of the multi-objective constraint calculation. If the current traffic density is high, the predicted speed can be appropriately reduced; otherwise, it can be appropriately increased. The generated optimal vehicle speed is displayed in real time on the vehicle's central control screen, helping the driver make more informed driving decisions. This series of steps ensures the effective implementation of the dynamic driving scenario topology rendering model and provides strong support for data visualization in electric vehicle intelligent dashcams. After completing the optimal speed prediction, a key step is to verify and adjust the prediction results through feedback. The accuracy of the speed prediction model can be adjusted and optimized in real time using feedback from actual driving. Experimental parameters are set, such as setting the feedback collection interval to once a minute, to ensure timely capture of changes in driving data. When implementing feedback adjustments, the actual vehicle speed is first monitored and compared with the model's predicted speed. Real-time vehicle speed data can be obtained through the OBD-II interface, and the actual speed and predicted speed are recorded. Next, the error between the actual speed and the predicted speed is analyzed. If the error exceeds a set threshold (e.g., ±5 km / h), the model parameters must be adjusted or retrained to improve its prediction accuracy. Using an incremental learning approach, new driving data is added to the model for retraining. Through continuous verification and adjustment, the optimal speed prediction model is continuously optimized to ensure its adaptability and accuracy under varying driving conditions. This process not only improves driving safety but also strengthens the dynamic adaptability of the vehicle's intelligent dashcam, providing users with a better driving experience.

[0017] Step S5: Calculate the multi-path probability of dynamic targets on the dynamic driving scene topology rendering model, perform credibility weight analysis, and extract the maximum probability behavior prediction trajectory of each target; In this embodiment, multi-path probability calculation is performed for dynamic targets within a topological rendering model of a dynamic driving scenario. By analyzing the motion state of each dynamic target, multiple possible behavior trajectories are generated, and a probability value is assigned to each trajectory. Experimental parameters are set, such as a 0.5-second time step for trajectory generation, to improve trajectory detail. When implementing multi-path probability calculation, the kinematic parameters of the dynamic target, such as velocity, acceleration, and steering frequency, are first obtained from the previous steps. These parameters are used to generate the target's predicted trajectory. A kinematic model (such as a uniform linear motion model or a uniformly accelerated motion model) can be used to calculate the possible trajectory of each target in the future time period. In specific implementation, a Monte Carlo method can be used to generate multiple random trajectories. Based on the current motion state and behavioral intent (such as acceleration, deceleration, steering, etc.), a certain random perturbation is set. By adding random noise at each time step, multiple different trajectories are generated. For example, if the target's current speed is 20 km / h, a random perturbation of ±5 km / h can be set to generate multiple possible trajectories. Then, by simulating the behavior of each trajectory in the future time period, the probability of its occurrence under specific conditions (such as road conditions and traffic density) is calculated. Using historical data and current environment characteristics, the posterior probability of each trajectory can be calculated based on Bayes' theorem. This process generates a set of possible behavioral trajectories and their corresponding probability values ​​for each dynamic target, forming a multi-path probability calculation result for the dynamic target. After the multi-path probability calculation is completed, a confidence weight analysis is performed to extract the most likely behavioral prediction trajectory for each target. By analyzing the probability of each trajectory, the most likely behavioral path is determined. Experimental parameters are set, such as setting a confidence analysis threshold of 70%, to ensure the reliability of the output results. When implementing the confidence weight analysis, the generated trajectories are first sorted. Each trajectory is sorted from high to low according to its probability value, ensuring that the most likely trajectory is at the top. This process can be implemented using the Python numpy.argsort() function. A threshold is set to filter out trajectories with probabilities above the threshold. For each dynamic target, the qualified trajectories and their probability values ​​are recorded. This ensures that only trajectories with high confidence are considered in practical applications. The system then further analyzes the context of each trajectory, including its relative position to other dynamic objects, road restrictions, traffic signals, and other factors, to ensure that the final selected trajectory is not only probabilistically advantageous but also logically consistent with the actual driving situation. The system extracts the trajectory with the highest probability of behavior prediction for each dynamic object and integrates it with other relevant information (such as object type and current speed) into a data structure for subsequent processing and visualization. After extracting the highest probability of behavior prediction trajectory, the results are organized into a structured data format. The analysis results are stored and displayed in a clear manner for subsequent visualization and decision support.Set the experiment parameters, such as setting the output format to JSON to facilitate interoperability with other systems. Build a dictionary or data frame containing information about each dynamic target. Each dynamic target entry should include the following fields: target ID, maximum probability trajectory coordinates, probability value, current velocity, acceleration, and so on. Convert the organized data frame to JSON format and serialize the data into a JSON string using the Python json library. This ensures data security and readability during transmission.

[0018] Step S6: Calculate the optimal safety distance for each target based on the maximum probability behavior prediction trajectory and the optimal vehicle speed, and perform visualization rendering on the vehicle's central control screen to perform visualization operations.

[0019] In this embodiment, the optimal safety distance is calculated for each target based on the predicted maximum probability behavior trajectory and the optimal vehicle speed. This step ensures that the vehicle maintains a safe distance from surrounding dynamic targets during driving, thereby improving driving safety. Experimental parameters are set, such as a reaction time of 2 seconds and a safety margin of 5 meters, to calculate a reasonable safety distance. To calculate the safety distance, the maximum probability behavior predicted trajectory and current speed of each dynamic target are first obtained. Based on the information extracted in the previous step, the trajectory data and speed parameters of each target are read. The optimal safety distance is then calculated using the following formula: Safety distance = reaction time × current speed + safety margin. If a target's current speed is 30 km / h (approximately 8.33 m / s) and the reaction time is 2 seconds, the safety distance is calculated as: Safety distance = 2 × 8.33 + 5 = 21.66 meters. In implementation, the numpy library in Python can be used for array calculations to batch process data for multiple dynamic targets. Furthermore, the safety margin should be appropriately adjusted to account for varying driving environments and target behaviors during the calculation. On urban roads, the safety margin might be set at 5 meters, while on highways, it might be increased to 10 meters. After the calculation is complete, the optimal safety distance information for each target is consolidated. The safety distance data is organized into a structured format for subsequent visualization. Experiment parameters are set, such as setting the output format to JSON for easy interaction with the vehicle's central control system. A dictionary or data frame is created containing the ID of each dynamic target, its predicted trajectory with the highest probability of its behavior, its current speed, and the calculated optimal safety distance. The data structure might look like this: After consolidating and outputting the safety distance information, visualization is performed on the vehicle's central control screen. The calculated safety distance and the dynamic target's behavior are presented to the driver in an intuitive manner to improve driving safety. Experiment parameters are set, such as setting the visualization update frequency to once per second to ensure timely information. When implementing visualization, the user interface (UI) must first be designed to ensure it effectively displays dynamic targets and their safety distances. The central control screen interface is constructed using front-end development technologies (such as HTML, CSS, and JavaScript), and visualization libraries (such as D3.js or Chart.js) are used to display the data. The implementation begins by mapping the positions of the vehicle and surrounding dynamic targets on the central control screen. For each target, its predicted trajectory with the highest probability is plotted, displaying its possible movement path. The calculated optimal safety distance is also marked around the target, using either a dotted line or a semi-transparent circle to indicate the safe distance range.

[0020] In this embodiment, refer to Figure 2 , is a flowchart of the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Capture full-process driving images based on intelligent driving recorders to obtain real-time video streams of driving scenes; Perform motion visual jitter elimination and compensation on the real-time video stream of the driving scene to construct a visual jitter optimized real-time video stream; Perform deep scene object recognition on real-time video streams optimized for visual jitter, perform segmentation processing, and mark multiple scene objects; Perform target semantic recognition on multiple scene targets to obtain scene target types; Calculate and track the speed, spatial position and topological interaction of multiple scene targets; Multi-level scene target feature perception is performed based on the scene target type, the speed, the spatial position and the topological interaction relationship, and a multi-level scene target information sequence is constructed.

[0021] In this embodiment, a vehicle-mounted camera is used for real-time monitoring. The camera should have a wide-angle lens to cover a wider range of driving scenarios. The dashcam's acquisition settings should be configured to ensure stable operation under various driving conditions, such as daytime, nighttime, and varying weather conditions. The dashcam should have real-time data transmission capabilities, transferring the captured video stream to the central processing unit (CPU) for subsequent processing. To ensure data integrity and efficiency, the dashcam should have intelligent storage management capabilities that automatically overwrite old data when storage space is insufficient. Furthermore, the dashcam should have collision detection capabilities to ensure that critical video data can be locked and saved in the event of an accident for subsequent analysis. After acquiring the real-time video stream, motion visual jitter removal compensation is performed. This eliminates jitter caused by vehicle motion and ensures video stream stability for subsequent analysis. Experimental parameters are set, such as setting the inter-frame difference threshold of the jitter compensation algorithm to 0.1, to ensure accurate compensation. When performing motion visual jitter removal, a video stabilization algorithm (such as an optical flow-based method or a feature point tracking algorithm) is first applied. By analyzing the motion differences between consecutive frames, jitter caused by vehicle motion is identified. Use the cv2.calcOpticalFlowFarneback function in the OpenCV library to calculate optical flow and obtain motion vectors for each frame. Use the calculated motion vectors to smooth the video frames. Each frame can be compensated for translation and rotation to ensure visual consistency. The resulting visually jitter-optimized live video stream significantly improves the accuracy of subsequent analysis. After video jitter compensation, perform deep scene object recognition and semantic segmentation on the visually jitter-optimized live video stream. Identify and label multiple scene objects in the video (such as pedestrians, vehicles, and traffic signs). Set experimental parameters, such as setting the input resolution of the object detection model to 640x480 to improve recognition speed and accuracy. Use a deep learning model (such as YOLOv5 or Mask R-CNN) for object recognition and segmentation. Input the optimized video frames into the trained model, which outputs bounding boxes and class labels for each object. During this step, ensure that the model is sufficiently trained to accurately recognize different types of scene objects. Next, perform semantic segmentation to separate the target region from the background. Using pixel-level classification methods, the specific location and shape of each object are marked. Using tools such as OpenCV and TensorFlow, the recognition and segmentation results are overlaid on the video frame to visually display the identified scene objects. The final result is a real-time video stream containing multiple labeled scene objects, providing basic data for subsequent target analysis. After completing target recognition and segmentation, target semantic recognition is performed on multiple scene objects to determine the scene object type. The specific type of each scene object (e.g., vehicle, pedestrian, non-motorized vehicle, etc.) is identified for further analysis.Set experimental parameters, such as setting the number of output categories of the semantic recognition model to 10 to cover major scene object types. Use a trained classification model (such as ResNet or VGG) to classify the identified objects. Input the image data of each object region into the classification model, which will return the class label and confidence score for each object. Ensure object recognition accuracy by selecting classification results with a confidence score above a set threshold (such as 0.7). Use the identified object types for feature extraction, recording each object's attributes (such as speed and spatial position). Continuously track each object in the video stream and update its state information. This process ensures a comprehensive understanding of scene objects and supports subsequent analysis of velocity and spatial interactions. After completing object semantic recognition, calculate and track the velocity, spatial position, and topological interactions of multiple scene objects. Analyze the dynamic relationships between objects in the scene. Set experimental parameters, such as setting the tracking window size to 5 frames, to ensure the continuity and accuracy of object states. Use a tracking algorithm (such as Kalman filtering or SORT) to track the motion state of each object. By analyzing the position changes of objects in consecutive frames, their velocity and acceleration are calculated. The spatial coordinate information of the objects is integrated to generate the motion trajectory of each object. Furthermore, the topological relationships between objects are analyzed to identify which objects are close to each other and calculate the distance between them. Clustering algorithms (such as DBSCAN) can be used to identify object groups and their relationships, further analyzing the dynamic changes of the scene. The output of this process is a dataset containing object velocities, positions, and topological relationships, which supports multi-level scene object feature perception. Multi-level scene object feature perception is performed based on scene object type, velocity, spatial position, and topological interactions, constructing a multi-level scene object information sequence. All extracted features are comprehensively analyzed to form a complete scene object information sequence. Experimental parameters are set, such as setting the information sequence time window to 1 second to ensure data timeliness. During this process, all object features obtained in the previous steps are integrated to form a multi-level information sequence. Each information sequence should include the object type, velocity, spatial position, and its interactions with other objects. These features are stored in a data structure (such as a dictionary or list) to facilitate subsequent querying and analysis. At the same time, visualization tools (such as Matplotlib or Plotly) are used to draw a scene target feature change graph to intuitively display the dynamic characteristics of different targets and their relationships.

[0022] In this embodiment, the specific steps of performing motion visual jitter elimination compensation on the real-time video stream of the driving scene to construct a visual jitter optimized real-time video stream are: Perform global image brightness optimization on the real-time video stream of the driving scene to obtain a globally brightness optimized video stream; Decompose the global brightness optimized video stream frame by frame to extract the time-series image frame sequence; Calculating a timestamp for each frame of the time-sequential image frame sequence; Obtain vehicle inertial sensor data; Dividing the vehicle inertial sensor data into time periods, performing data matching based on the timestamp of each frame, and extracting the vehicle inertial sensor parameters corresponding to the timestamp; Calculate the relative rotation and displacement between multiple frames based on the vehicle inertial sensor parameters; Perform pixel-level motion optical flow estimation on the temporal image frame sequence to obtain pixel-level motion optical flow information; Motion visual jitter elimination and compensation are performed on a temporal image frame sequence based on pixel-level motion optical flow information, the relative rotation and displacement, so as to construct a visual jitter optimized real-time video stream.

[0023] In this embodiment, the global image brightness optimization is performed on the real-time video stream of the driving scene to obtain a global brightness optimized video stream. Improve the video quality to ensure that the driving scene can be clearly displayed under various lighting conditions. Set the experimental parameters, such as setting the target brightness value of the brightness optimization to 1.2 times the average brightness of the original image to enhance the visibility of the image. When implementing image brightness optimization, use an image processing library (such as OpenCV) to process each frame of the image. Calculate the average brightness of the current frame, convert the image to a grayscale image through cv2.cvtColor, and then use cv2.mean to calculate the average brightness value. Then, according to the set target brightness value, calculate the required brightness gain. Use the gamma correction method to adjust the brightness of the image. The specific implementation is to convert each pixel value into a new value: new value = 255× ; where γ is the set gain parameter. The resulting globally brightness-optimized video stream ensures improved video quality under various lighting conditions, providing a clear image foundation for subsequent processing. After acquiring the globally brightness-optimized video stream, perform frame-by-frame image decomposition to extract a time-series image frame sequence. Decompose the video stream into continuous image frames for subsequent analysis. Set experimental parameters, such as setting an image frame extraction interval of 30 frames per second, to ensure sufficient time-series information. Use the cv2.VideoCapture function in the OpenCV library to read the video stream and extract images frame by frame in a loop. After each frame is read, store it in a list. To ensure continuous extraction of image frames, use the cv2.waitKey function to control the extraction rate and ensure synchronization with the video stream frame rate. During the extraction process, image format conversion is also required to facilitate subsequent processing. Convert each frame to a suitable format (such as RGB or grayscale) and ensure that each frame has consistent dimensions. The resulting time-series image frame sequence provides the basis for subsequent timestamp calculation and data matching. After extracting the time-series image frame sequence, calculate the timestamp for each frame. Assign accurate time information to each frame to facilitate subsequent matching with the vehicle's inertial sensor data. Set experimental parameters, such as setting the timestamp accuracy to milliseconds (1 millisecond). When calculating the timestamp, first record the start time of the video stream. Use the Python time library to obtain the current time, which serves as the reference time for the start of the video. During the frame-by-frame extraction process, assign a timestamp to each frame. The timestamp is calculated as: timestamp = start time + (current frame index / frame rate). This method generates an accurate timestamp for each frame, ensuring accurate matching in subsequent data matching. After calculating the timestamp, acquire the vehicle's inertial sensor data. Collect real-time data on the vehicle's motion status to facilitate subsequent time segmentation and data matching. Set experimental parameters, such as setting the sampling frequency to 100 Hz, to ensure sufficient sensor data. Vehicle inertial sensors (such as accelerometers and gyroscopes) typically transmit data via the CAN bus or other communication protocols. Use an appropriate interface (such as UART or I2C) to read sensor data and store it as time series data. Each data point should include a timestamp, acceleration (three-axis data), and angular velocity (three-axis data). During data collection, ensure real-time monitoring of sensor operating status and handle any anomalies (such as sensor failure or data loss). The resulting vehicle inertial sensor data will provide the necessary information for subsequent time period segmentation and data matching.

[0024] The vehicle inertial sensor data is divided into time periods, and data matching is performed based on the timestamp of each frame to extract the vehicle inertial sensor parameters corresponding to the timestamp. The image data is then associated with the sensor data for subsequent motion analysis. Experimental parameters are set, for example, to a time period of 1 second. When performing time period division, the sensor data is first segmented based on the timestamp. The Python pandas library can be used to convert the sensor data into a DataFrame format, grouping them by timestamp, with each group containing data within 1 second. During data matching, the sensor parameters for the corresponding time period are found by iterating through the timestamps of each frame. The merge_asof function can be used to achieve approximate matching, matching the timestamps of the image frame with the timestamps of the sensor data to ensure that relevant sensor data is captured for each frame. The resulting matching data provides basic information for subsequent relative rotation and displacement calculations, ensuring accurate analysis. Relative rotation and displacement between multiple frames are calculated based on the vehicle inertial sensor parameters. Changes in the vehicle's motion are analyzed to ensure accurate compensation of the time-series image frames. Experimental parameters are set, for example, to set the units for rotation calculations to degrees and the units for displacement calculations to meters. When calculating relative displacement, the acceleration data from the inertial sensor is integrated to calculate the displacement between each frame. Displacement can be calculated using the following formula: Displacement = ∫a(t)dt; where a(t) is the acceleration data. By numerically integrating the acceleration data, the displacement within each time period is obtained. To calculate the rotation angle, the angular velocity data provided by the gyroscope is similarly integrated to obtain the rotation angle between each frame. By converting the angle to radians and performing the calculation, relative rotation information for each image frame is generated. The resulting relative rotation and displacement information provides the necessary data for subsequent motion visual jitter compensation. After the relative rotation and displacement calculations are completed, pixel-level motion optical flow is estimated for the time-series image frames to obtain pixel-level motion optical flow information. Motion changes between image frames are analyzed for jitter compensation. Experimental parameters are set, such as setting the window size for the optical flow calculation to 5x5 pixels, to ensure estimation accuracy.

[0025] Use the cv2.calcOpticalFlowFarneback function in the OpenCV library to calculate optical flow for consecutive frames. This function calculates optical flow based on a pyramidal method, analyzing pixel motion to obtain optical flow vectors. Input each image frame into the function and set appropriate parameters (such as the number of pyramid levels and window size). During the calculation process, ensure that the optical flow results are post-processed to filter out unreasonable optical flow values ​​(such as those exceeding a set threshold). The resulting pixel-level motion optical flow information provides the necessary basic data for subsequent motion judder compensation. Based on this pixel-level motion optical flow information, relative rotation, and displacement, motion judder compensation is performed on a time-series image frame sequence to construct a visually judder-optimized real-time video stream. By compensating for judder, the stability and clarity of the video stream are improved. Experimental parameters are set, such as limiting the reconstruction accuracy to within 2 pixels. During motion judder compensation, each frame is transformed based on the optical flow information, relative rotation, and displacement data. Using an affine transformation or perspective transformation model, each frame is adjusted to the calculated motion information. The specific method is as follows: Calculate the transformation matrix: Construct an affine transformation matrix based on the optical flow and rotation and displacement information. Apply the transformation: Use the cv2.warpAffine or cv2.warpPerspective function to transform the image to obtain the compensated image. Overlay the frame: Overlay the compensated image with the original image to generate the final visually jitter-optimized video stream.

[0026] In this embodiment, refer to Figure 3 , is a flowchart of the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Intelligently identify road boundary elements based on visual jitter optimization of real-time video streams and extract road boundary element information; Calculate slope changes and road curvature based on visual jitter optimization of real-time video streams to obtain road condition information; Based on the multi-level scene target information sequence, the road boundary element information, slope changes and road curvature are mined in real time to construct the driving scene road topology map; Optimize real-time video streams based on visual jitter to track dynamic targets and road topology, and extract real-time changing trajectories of dynamic targets and road topology; Based on the real-time change trajectory, multi-dimensional scene topology trend evolution fitting is performed to construct the scene motion trend matrix; Based on the scene motion trend matrix, the driving scene road topology map is rendered in spatiotemporal connection and the scene evolution is smoothed to construct a dynamic driving scene topology rendering model.

[0027] In this embodiment, intelligent road boundary feature recognition is performed based on a real-time video stream optimized for visual jitter to extract road boundary feature information. Accurately identifying and extracting road boundary features provides fundamental data for subsequent road analysis. Experimental parameters are set, such as a threshold of 0.5 for boundary detection, to ensure detection accuracy. Image processing and computer vision techniques are used to implement road boundary recognition. The optimized video stream is converted into a grayscale image for ease of subsequent processing. Next, edge detection algorithms (such as Canny edge detection) are used to extract edge information from the image, and an edge map is calculated using the cv2.Canny function. Subsequently, a Hough transform is applied to detect lines to identify road boundaries. By setting appropriate parameters (such as minimum length and threshold), key road boundary lines can be extracted. The identified boundary information is overlaid back onto the video frame to form a visual boundary feature image, ensuring that subsequent analysis is based on accurate boundary information. After successfully extracting road boundary features, the real-time video stream optimized for visual jitter is used to calculate slope changes and road curvature to obtain road surface condition information. The road's geometric characteristics are analyzed to provide a basis for driving safety. Set the experimental parameters, such as setting the window size for curvature calculation to 5 frames to ensure the stability of the calculation. When calculating the slope, first calculate the inclination angle of the road by extracting the road boundary features. You can use the linear regression method to fit the road boundary line to obtain the slope of the road. Use the numpy.polyfit function to fit the road boundary points and calculate its slope value. For the calculation of curvature, use the second-order derivative of the curve to represent the curvature change of the road. The curvature formula is: ; Where y^'' and y^' are the second and first derivatives of the road boundary, respectively. By calculating the curvature value of each frame, a curvature change sequence is generated to ensure that the changes in the road can be observed on the time axis. By calculating the slope and curvature, road condition information is generated to form a data set that records the slope and curvature values ​​of each frame to provide support for subsequent road topology mining. Based on the multi-level scene target information sequence, real-time road topology mining is performed on road boundary feature information, slope changes, and road curvature to construct a driving scene road topology map. The extracted boundary information and geometric features are integrated into a topological model for subsequent analysis and application. Experimental parameters are set, such as setting the time window for topology mining to 10 seconds to ensure the timeliness of the data. When performing topology mining, the extracted boundary feature information, slope changes, and curvature data are first integrated. Using graph theory methods, the boundary lines of the road are regarded as nodes, and the slope and curvature changes are regarded as edge features. By establishing a graph structure, the connection relationship between each node is recorded. Use network analysis tools (such as NetworkX) to construct a road topology map, ensuring that the connectivity of each node accurately reflects the road's geometric characteristics. Visualization tools are also used to display the topology map, visually illustrating the road's structural characteristics. The resulting road topology map for the driving scenario provides a foundational model for dynamic object tracking and scene evolution analysis. After constructing the road topology map, dynamic objects and road topology are tracked dynamically within a real-time video stream optimized for visual jitter. This allows for the extraction of real-time trajectories of these objects and their relationship to the road topology. Dynamic objects in the scene and their relationship to the road topology are monitored in real time. Experimental parameters are set, such as a detection frequency of 10 times per second for dynamic object tracking, to ensure real-time performance. Object detection and tracking algorithms (such as YOLO or SORT) are used to detect and track dynamic objects in the video stream. Each time a dynamic object is detected, its position in the video frame and the corresponding timestamp are recorded. The constructed road topology map is then used to analyze the relationship between the dynamic object and the road topology. By calculating the distance between the dynamic object and the road node, the road segment on which the object travels is identified, and the corresponding topological information is recorded. A data structure can be used to store the trajectory information of each dynamic object, including its position, velocity, and corresponding road topology node. The resulting dynamic target trajectory data will provide the necessary information support for subsequent scene topology trend evolution fitting. Based on the real-time changing trajectory, multi-dimensional scene topology trend evolution fitting is performed to construct a scene motion trend matrix. The evolution trend of dynamic targets in the road topology is analyzed to facilitate subsequent visualization. Experimental parameters are set, such as setting the fitting model to a polynomial regression model, to ensure flexibility in data fitting. When performing trend evolution fitting, the trajectory information of all dynamic targets is first collected to construct a dataset containing time, position, speed, and road topology nodes. Then, the polynomial regression model is applied to fit the trajectory of each dynamic target to generate a motion trend matrix.Polynomial regression fitting can be performed using the PolynomialFeatures and LinearRegression functions in the Scikit-learn library. By analyzing the trajectories of multiple dynamic targets, a comprehensive scene motion trend matrix is ​​constructed, recording the motion characteristics and changing trends of different targets. The resulting motion trend matrix provides basic data for subsequent scene topology rendering, ensuring accurate reflection of the dynamic scene evolution. Based on the scene motion trend matrix, the driving scene road topology map is rendered spatiotemporally and smoothly, thereby constructing a dynamic driving scene topology rendering model. All dynamic information is integrated into a visualization model for real-time display. Experimental parameters are set, such as a rendering frame rate of 30 fps, to ensure smooth dynamic effects. During spatiotemporal rendering, the scene motion trend matrix is ​​combined with the road topology map using visualization tools such as Matplotlib or Plotly. Based on the motion trajectories of the dynamic targets, the real-time changes of the targets within the road topology are plotted. Different colors and shapes of markers can be used to represent different types of dynamic targets. At the same time, a smoothing algorithm (such as the Savitzky-Golay filter) is applied to smooth the dynamic changes in the scene, reducing the impact of noise on the rendering effect. By smoothing the dynamic target trajectory, the final rendering effect is more natural and realistic. The resulting dynamic driving scene topology rendering model will be able to display dynamic changes in the driving scene in real time, supporting driving safety monitoring and intelligent decision-making. This series of steps ensures comprehensive analysis and visualization of the driving scene, providing an effective data processing solution for the intelligent driving recorder system of electric vehicles.

[0028] In this embodiment, refer to Figure 4 , is a flowchart of the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include: Obtain vehicle operating status parameters; calculate real-time vehicle speed, engine power output, vehicle battery status, and motor temperature based on the vehicle operating status parameters to obtain multiple vehicle operating indicators; Conduct multi-parameter correlation analysis on multiple vehicle operation indicators and extract correlation patterns of multiple indicators; Identify real-time driving behavior data based on vehicle operating status parameters; Perform vehicle transient response analysis on multiple vehicle operating indicators based on real-time driving behavior data to obtain real-time vehicle response data; Based on the correlation rules of multiple indicators, a comprehensive status correlation evaluation is performed on the vehicle status response data to build a real-time vehicle status evaluation report.

[0029] In this embodiment, when implementing data collection, the vehicle's operating status parameters are generally obtained through the on-board diagnostic system (OBD). An OBD-II adapter is used to connect to the vehicle's OBD interface to read the vehicle's real-time data, including speed, engine speed, throttle opening, etc. A Python library (such as pyOBD or python-OBD) can be used to communicate with the OBD-II adapter to obtain real-time data. During the data acquisition process, ensure that the connection status and data transmission quality are monitored to avoid data loss due to failures. The acquired parameters are stored in a data structure (such as a dictionary or data frame) for subsequent processing and analysis. The resulting vehicle operating status parameters will provide basic data for subsequent indicator calculation and analysis.

[0030] Through the OBD-II interface, we read the following parameters in real time: Current speed: 60 km / h Engine speed: 3000 RPM Throttle opening: 30% Engine temperature: 85°C Vehicle battery voltage: 12.5 V Using a program (e.g., Python), read this data via the CAN bus and store it in a data structure (e.g., a dictionary or data frame) for subsequent processing. Real-time vehicle speed can be directly read using speed parameters obtained via the OBD. If more accurate results are required, correction can be performed using speed data obtained from the GPS module. Engine power output can be calculated using the following formula: Power = (Torque × Speed) / 5252; both torque and speed are obtained via the OBD. The calculation results are stored in a data structure, forming a single indicator dataset containing real-time vehicle speed and power output. For vehicle battery status, battery voltage and current data can typically be obtained from the battery management system (BMS) to calculate the battery's State of Charge (SOC) and State of Health (SOH). Motor temperature should also be obtained using appropriate sensors. After obtaining multiple vehicle operating indicators, perform multi-parameter correlation analysis to identify correlation patterns among these indicators. Identify correlations between different indicators to provide a basis for subsequent driving behavior identification and response analysis. Set experimental parameters, such as setting a 95% confidence level for the correlation analysis. Use statistical analysis tools (e.g., Python's Pandas and SciPy libraries) to perform correlation analysis on multiple indicators. Organize all indicators into a data frame format, ensuring that data for each indicator within the same time period is aligned. Use the pandas.DataFrame.corr() method to calculate the correlation coefficient matrix between each indicator and obtain the linear correlation between them. Use visualization tools (such as Seaborn) to create heatmaps to visually display the correlations between different indicators. During the analysis, pay special attention to combinations of indicators with significant correlations and record these correlation patterns. This process will provide the necessary basis for identifying real-time driving behavior data based on vehicle operating status parameters. Identify real-time driving behavior data based on vehicle operating status parameters. Determine the driver's driving behavior characteristics by analyzing vehicle operating status. Set experimental parameters, such as using a decision tree classification model for driving behavior identification, to ensure classification accuracy. When implementing driving behavior identification, first define common driving behavior types (such as sudden acceleration, sudden braking, constant speed driving, and turning). Then, use a machine learning model to classify the acquired vehicle operating status parameters. Use the scikit-learn library to build a decision tree model and train it on driving behaviors. During training, use labeled historical data as the training set to ensure the model can accurately identify different driving behaviors. Once model training is complete, real-time vehicle state parameters are fed into the model to identify real-time driving behaviors. The resulting real-time driving behavior data provides crucial information for subsequent vehicle transient response analysis. Based on this real-time driving behavior data, a transient response analysis is performed on multiple vehicle operating indicators to generate real-time vehicle response data. This allows the vehicle's dynamic response performance to different driving behaviors to be evaluated.Experimental parameters are set, such as a 5-second time window for response analysis to capture transient changes. During transient response analysis, the current driving behavior is first identified, followed by a time series analysis of relevant vehicle operating indicators. The vehicle's response speed and stability are assessed by calculating the change in each indicator after the driving behavior begins. Time series analysis methods (such as moving average or exponential smoothing) are used to smooth the data and extract key response characteristics (such as response time and peak variation). These characteristics help understand the vehicle's performance under specific driving behaviors. The resulting real-time vehicle response data provides foundational information for subsequent state correlation assessments. Based on the correlation patterns of multiple indicators, a comprehensive state correlation assessment is performed on the vehicle state response data to construct a real-time vehicle state assessment report. All analysis results are integrated into a comprehensive vehicle state assessment report. Experimental parameters are set, such as setting the report update frequency to once a minute, to ensure data timeliness. During the state correlation assessment, real-time driving behavior data, vehicle operating indicators, and transient response data are integrated into a comprehensive data frame. A weighted average method is used to comprehensively score the different indicators to ensure that each is appropriately represented in the assessment. Data visualization tools are also used to generate charts showing vehicle status trends and key performance indicators. These charts help users intuitively understand the vehicle's current performance and potential issues. The resulting real-time vehicle status assessment report provides drivers and managers with the necessary information to assess vehicle performance and make appropriate decisions. This series of steps ensures efficient data collection and analysis within the electric vehicle intelligent driving recorder system, providing a scientific basis for safe driving and performance optimization of electric vehicles. In this embodiment, step S4 includes the following steps: Identify road speed limit signs and regional speed limit regulations based on a dynamic driving scene topology rendering model to obtain road time limit information; Real-time dynamic target recognition is performed based on the dynamic driving scene topology rendering model, and road traffic density is calculated to obtain road traffic density characteristics; Carry out vehicle traffic distribution mining based on road traffic density characteristics and generate vehicle traffic density distribution characteristics; Identify the distribution of road obstacles and changes in road curvature based on the dynamic driving scene topology rendering model, and generate road complexity data; Conduct road safety assessment based on vehicle traffic density distribution characteristics and road complexity data to obtain road safety assessment results; Based on the real-time vehicle status assessment report, multi-objective constraint calculations are performed on the road safety assessment results, and the optimal vehicle speed is predicted to obtain the optimal vehicle speed.

[0031] In this embodiment, the recognition process uses computer vision technology, specifically deep learning models (such as YOLO or SSD), to detect and identify speed limit signs. A video stream of a dynamic driving scene is fed into a trained object detection model, which identifies all speed limit signs and assigns them category labels. The location information of the identified speed limit signs is extracted. The recognition results can be post-processed using OpenCV to extract the sign coordinates and speed limit values. The identified speed limit information is stored in a data structure for subsequent use. After recognition, the generated road time limit information is displayed in real time on the driver interface to help drivers comply with speed limit regulations. Furthermore, speed limit information is used for subsequent traffic density analysis and safety assessments. After identifying speed limit signs, real-time dynamic object recognition is performed based on a dynamic driving scene topology rendering model, and road traffic density is calculated to obtain road traffic density features. Dynamic objects (such as vehicles and pedestrians) on the road are monitored, and traffic density is calculated. Experimental parameters are set, for example, to a dynamic object detection frequency of 10 times per second to ensure real-time performance. Dynamic objects in the video stream are detected using the same deep learning model used for speed limit sign recognition. The model identifies all dynamic objects and provides their location, velocity, and category labels. The video stream is captured using cv2.VideoCapture and processed frame by frame to ensure object detection in every frame. After dynamic object detection is complete, traffic density features are calculated. Traffic density is calculated by counting the number of dynamic objects detected in each frame and combining it with road width information. The formula is: traffic density = number of dynamic objects / road width. The calculated traffic density features are stored in a time series data structure for subsequent analysis and visualization. After obtaining road traffic density features, vehicle traffic distribution mining is performed based on these features to generate vehicle traffic density distribution features. Traffic flow in different areas is analyzed to identify high- and low-density areas. Experimental parameters are set, such as a 5-minute analysis window, to ensure data stability. Traffic density data within each time period is clustered (e.g., using K-means clustering) to identify traffic distribution features for different areas. The traffic density data is converted into points in a two-dimensional coordinate system, representing the traffic density value for each area. These points are clustered using the K-means algorithm, with the number of clusters set to 3 (low density, medium density, and high density). Cluster analysis is performed using K-means in the scikit-learn library to obtain cluster centers and the areas corresponding to each center. The resulting vehicle traffic density distribution features will help identify road congestion and provide a basis for subsequent road complexity analysis and safety assessment. Based on the dynamic driving scenario topology rendering model, the distribution of road obstacles and changes in road curvature are identified to generate road complexity data. Road safety and drivability are assessed. Experimental parameters are set, for example, setting a target obstacle detection accuracy of 90%.A deep learning model (such as Faster R-CNN) is used to detect road obstacles. This model identifies both static and dynamic obstacles on the road (such as other vehicles, pedestrians, and traffic cones). By extracting the location and category of obstacles, the obstacle data is stored in a data structure. Simultaneously, changes in road curvature are analyzed. Based on the previously extracted road boundary information, a curvature calculation formula is used to analyze road changes. Curvature values ​​are calculated for each segment and combined with obstacle data to generate road complexity data. This resulting road complexity data provides important information for road safety assessment, ensuring effective identification of potential hazards during driving. After obtaining vehicle traffic density distribution characteristics and road complexity data, a road safety assessment is conducted on this data to generate road safety results. Traffic density and road complexity are comprehensively analyzed to assess road safety. Experimental parameters are set, such as using a multivariate linear regression model to ensure accuracy. A multivariate linear regression model is used to establish a safety assessment model, using traffic density and road complexity as independent variables. By collecting historical accident data, a model is trained to identify accident rates under varying density and complexity conditions. Once the model is trained, real-time traffic density and complexity data is fed into the model to calculate a road safety score. The score is displayed on the driver interface, along with corresponding safety recommendations. The resulting road safety assessment provides drivers with important safety guidance, helping them make better decisions while driving.

[0032] In this embodiment, step S5 includes the following steps: Accurately segment dynamic targets in the topological rendering model of dynamic driving scenes and extract each dynamic target; Calculate the velocity vector, acceleration change rate and direction turning frequency of each dynamic target to obtain the kinematic parameters of each target; Perform intention reasoning and behavior prediction based on the kinematic parameters to generate a set of behavior prediction trajectories for each target point; Multi-path probability calculation is performed on the behavior prediction trajectory set, and credibility weight analysis is performed to extract the maximum probability behavior prediction trajectory of each target.

[0033] In this embodiment, the dynamic driving scene topology rendering model is used to accurately segment dynamic targets to extract each dynamic target. Each dynamic target in the scene (such as vehicles, pedestrians, etc.) is accurately identified and segmented to provide basic data for subsequent kinematic parameter calculations. Experimental parameters are set, for example, the accuracy target of the segmentation network is set to 95%. When implementing dynamic target segmentation, deep learning semantic segmentation technology, such as U-Net or Mask R-CNN, is used. The video stream of the dynamic driving scene is input into the trained segmentation model. The model will process each frame of the image and generate a segmentation mask for each dynamic target. By processing each frame of the image, the model will output the pixel-level segmentation result of each dynamic target. The segmentation results are post-processed using OpenCV to eliminate noise and small isolated areas to ensure the accuracy of the segmentation results. The segmentation results of each dynamic target are stored in a data structure, including the target's location information and category label. The dynamic target segmentation data finally generated will provide clear basic information for subsequent kinematic parameter calculations, ensuring that each target can be accurately identified and analyzed. After successfully extracting the dynamic target, calculate the velocity vector, acceleration rate, and direction turning frequency of each dynamic target to obtain the kinematic parameters of each target. Analyze the motion state of each dynamic target to provide a basis for behavior prediction. Set the experimental parameters, such as setting the time interval for velocity calculation to 1 second to ensure the timeliness of the data. Use a target detection algorithm (such as a Kalman filter) to track each dynamic target. By matching the target's position information between consecutive frames, calculate the velocity vector of each frame. The calculation formula for the velocity vector is: Where Δx and Δy represent the target's position change within the time interval Δt. Calculate the rate of change of acceleration for each target. The formula for calculating acceleration is: acceleration = Δv / Δt; where Δv is the change in velocity. Acceleration is obtained by calculating the velocity of two consecutive frames. The frequency of turning can be calculated by analyzing the target's trajectory. Set a threshold (e.g., 30 degrees) to determine whether the target has turned, and record the frequency of turning. Intention inference and behavior prediction are performed based on the kinematic parameters, generating a set of predicted behavior trajectories for each target. By analyzing the target's motion state, its possible future trajectory is predicted. Experimental parameters are set, such as a prediction time range of 5 seconds, to ensure the accuracy of trajectory prediction. To perform intention inference, the kinematic parameters of each target are first analyzed. Combined with historical trajectory data, a machine learning model (such as a random forest or LSTM) is used to perform intention inference. The model inputs parameters such as the target's velocity, acceleration, and turning frequency, and outputs a target intent classification (e.g., acceleration, deceleration, turning, etc.). Next, after the intent is determined, a set of predicted behavior trajectories is generated for each target. Based on the target's current state and intention, a kinematic model (such as uniform linear motion or uniformly accelerated motion) is used to generate multiple possible trajectories. Random trajectories can be generated based on Monte Carlo methods to ensure that the diversity of target behavior is accounted for. The generated trajectory data is collected to form a set of predicted behavior trajectories for each dynamic target, containing different possible paths and corresponding time information. Multi-path probability calculation and credibility weight analysis are performed on this set of predicted behavior trajectories to extract the most probable behavior prediction trajectory for each target. This probabilistic analysis determines the most probable path for the target's behavior to improve prediction accuracy. Experimental parameters are set, such as setting the number of samples for probability calculation to 1000, to ensure the reliability of the results. During the multi-path probability calculation, each predicted trajectory is first evaluated to calculate its probability of occurrence. A probability value can be assigned to each trajectory based on historical data and current environmental factors (such as traffic density and road type). A Bayesian network is used to integrate environmental factors with the predicted trajectory to calculate the posterior probability of each trajectory. Next, a credibility weight analysis is performed. Using a weighted average method, the probability of each trajectory is combined with its corresponding weight to extract the trajectory with the highest probability. The weighted average can be calculated using the np.average() function in the numpy library. The resulting maximum probability behavior prediction trajectory provides the most reliable prediction of the dynamic target's future behavior, helping the driver or the autonomous driving system make appropriate decisions. This series of steps ensures in-depth analysis of target behavior in dynamic driving scenarios and provides strong support for visualization methods for electric vehicle intelligent driving recorders.

[0034] In this embodiment, step S6 includes the following steps: Based on the dynamic driving scene topology rendering model, real-time road topology state evolution recognition is performed to extract dynamic road topology state evolution features; Performing a road driving safety risk analysis based on the dynamic road topology state evolution characteristics and the maximum probability behavior prediction trajectory to generate a road driving risk level; Calculate the optimal safety distance for each target based on the road driving risk level and the optimal vehicle speed to obtain the optimal safety distance for each target; Performing real-time dynamic mapping on the dynamic driving scene topology rendering model according to the optimal safety distance to construct a holographic driving scene model; Perform visual rendering of the vehicle's central control screen based on the holographic driving scene model to perform visualization operations.

[0035] In this embodiment, a dynamic driving scenario topology rendering model is used to identify real-time road topology state evolution and extract dynamic road topology state evolution features. This allows for real-time monitoring and identification of road state changes to facilitate safety risk analysis. In the implementation, sensors (such as lidar and cameras) are used to acquire real-time road topology data. Data fusion technology is used to integrate data from various sources to generate a current road topology model.

[0036] Assume that at a certain moment, the road topology information captured by the data is as follows: Road width changes: from 3.5m to 2.5m; Road curvature change: curvature increased from 0.01 / m to 0.03 / m; Road conditions: slippery; Use machine learning algorithms (such as convolutional neural networks (CNN)) to analyze these topological data and extract the dynamic road topology state evolution characteristics. Identify the change patterns of road states, such as "narrowing" and "increasing curvature", by training the model. Successfully identify the dynamic road topology state evolution characteristics and provide basic data for subsequent safety risk analysis. After extracting the dynamic road topology state evolution characteristics, perform road driving safety risk analysis based on these characteristics and the maximum probability behavior prediction trajectory to generate a road driving risk level. Assess the safety of the current road driving environment. Set a risk assessment model based on the road topology state evolution characteristics and the maximum probability behavior prediction trajectory of the dynamic target. The risk level can be analyzed through logistic regression or fuzzy logic systems. Set the following risk assessment criteria: The risk level increases when the road curvature increases and the vehicle speed exceeds a certain value (such as 60 km / h).

[0037] Slippery road conditions increase the risk factor.

[0038] Assume that according to the analysis, the risk factor of the current road is calculated as follows: Curvature change risk factor: 2 (medium risk) Road surface condition risk factor: 3 (high risk) Calculate the overall risk level based on the set weights: Overall risk level = Curvature change risk factor + Road surface condition risk factor = 2 + 3 = 5 (high risk). The road driving risk level is successfully generated, providing a basis for subsequent safe distance calculations. After obtaining the road driving risk level, the optimal safe distance is calculated for each target based on the level and the optimal vehicle speed to determine the optimal safe distance for each target. This ensures safe spacing between targets and reduces accident risk. The safe distance calculation formula is: Optimal safe distance = Vehicle speed × Time interval. Assume the current optimal speed is 50 km / h, and the time interval is set to 2 seconds based on the road risk level. First, convert the speed to meters per second: Speed ​​= 50 km / h / 3.6 ≈ 13.89 m / s. Calculate the optimal safe distance using the formula: Optimal safe distance = 13.89 m / s × 2 s = 27.78 m. Repeat this calculation for each dynamic target, combining the target's motion state and road risk level to determine the optimal safe distance for each target. Assume the optimal safe distance for target 1 is 28 m, and for target 2 is 30 m. After determining the optimal safe distance for each target, the dynamic driving scene topology rendering model is dynamically mapped in real time based on these safe distances to construct a holographic driving scene model. This provides the driver with intuitive information about the driving environment. Real-time data is used to update every element in the topology rendering model, including the road, dynamic targets, and their safe distances. The holographic driving scene model is created using 3D modeling software (such as Unity or Blender), rendering the road status and dynamic targets in real time.

[0039] Assume that in the model, the distance relationship between the dynamic target and the road is as follows: Target 1 (car): The optimal safety distance from the road is 28 m; Target 2 (pedestrians): The optimal safety distance from the road is 30 m; Visualize this distance information and build a safe area for each target in the dynamic scene. Use different colors or transparency to represent the safe area to facilitate quick identification by the driver. Successfully build a holographic driving scene model to provide data support for subsequent visualization rendering. Perform visualization rendering on the vehicle's central control screen based on the holographic driving scene model to perform visualization operations. Display real-time driving information and safety tips on the driver's central control screen. Design a visualization interface to display the rendering results of the holographic model through a graphical user interface (GUI). This can be implemented using Unity, Qt, or other GUI development tools. In the interface, display key information such as the target's motion status, optimal safety distance, road risk level, etc. For example: Target 1: Speed ​​= 50 km / h, Safety distance = 28 m, Risk level = High; Target 2: Speed ​​= 10 km / h, Safety distance = 30 m, Risk level = Medium; Through dynamic updates, the data in the interface is synchronized with the real-time status of the vehicle, improving the driver's safety awareness. The successful implementation of visual rendering of the vehicle's central control screen provides the driver with real-time driving information support and improves driving safety.

[0040] In this embodiment, a data visualization system for an electric vehicle intelligent driving recorder is provided, which is used to execute the data visualization method for an electric vehicle intelligent driving recorder as described above, including: The video processing module is used to obtain real-time video streams of driving scenes based on intelligent driving recorders, perform motion visual jitter elimination and compensation, and perceive multi-level scene target features to construct a multi-level scene target information sequence; The scene topology rendering module is used to perform spatiotemporal connection rendering based on multi-level scene target information sequences and smooth scene evolution, thereby constructing a dynamic driving scene topology rendering model; The vehicle status assessment module is used to obtain vehicle operating status parameters, perform multi-parameter correlation analysis and vehicle transient response analysis, and construct a real-time vehicle status assessment report; A multi-objective constraint module is used to perform multi-objective constraint calculations based on a dynamic driving scenario topology rendering model and to predict the optimal vehicle speed based on a real-time vehicle status assessment report to obtain the optimal vehicle speed. The behavior trajectory prediction module is used to calculate the multi-path probability of dynamic targets in the dynamic driving scene topology rendering model, perform credibility weight analysis, and extract the maximum probability behavior prediction trajectory for each target; The visualization rendering module is used to calculate the optimal safety distance for each target based on the maximum probability behavior prediction trajectory and the optimal vehicle speed, and perform visualization rendering on the vehicle's central control screen to perform visualization operations.

[0041] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0042] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily 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 is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A data visualization method for an electric vehicle intelligent driving recorder, characterized in that: The following steps are involved: Step S1: Acquire a real-time video stream of the driving scene based on the intelligent driving recorder, perform motion visual jitter elimination compensation and multi-level scene target feature perception, and construct a multi-level scene target information sequence; Step S2: Perform spatiotemporal connection rendering based on the multi-level scene target information sequence and perform scene evolution smoothing to construct a dynamic driving scene topology rendering model; Step S3: Obtain vehicle operating status parameters, perform multi-parameter correlation analysis and vehicle transient response analysis, and construct a real-time vehicle status assessment report; Step S4: performing multi-objective constraint calculation based on the dynamic driving scene topology rendering model, and performing optimal vehicle speed prediction based on the real-time vehicle status assessment report to obtain the optimal vehicle speed; Step S5: Calculate the multi-path probability of dynamic targets on the dynamic driving scene topology rendering model, perform credibility weight analysis, and extract the maximum probability behavior prediction trajectory of each target; Step S6: Calculate the optimal safety distance for each target based on the maximum probability behavior prediction trajectory and the optimal vehicle speed, and perform visualization rendering on the vehicle's central control screen to perform visualization operations.

2. The data visualization method of the electric vehicle intelligent driving recorder according to claim 1 is characterized in that: The specific steps of step S1 are: Capture full-process driving images based on intelligent driving recorders to obtain real-time video streams of driving scenes; Perform motion visual jitter elimination and compensation on the real-time video stream of the driving scene to construct a visual jitter optimized real-time video stream; Perform deep scene object recognition on real-time video streams optimized for visual jitter, perform segmentation processing, and mark multiple scene objects; Perform target semantic recognition on multiple scene targets to obtain scene target types; Calculate and track the speed, spatial position and topological interaction of multiple scene targets; Multi-level scene target feature perception is performed based on the scene target type, the speed, the spatial position and the topological interaction relationship, and a multi-level scene target information sequence is constructed.

3. The data visualization method of the electric vehicle intelligent driving recorder according to claim 2 is characterized in that: The specific steps of performing motion visual jitter elimination compensation on the real-time video stream of the driving scene to construct a visual jitter optimized real-time video stream are as follows: Perform global image brightness optimization on the real-time video stream of the driving scene to obtain a globally brightness optimized video stream; Decompose the global brightness optimized video stream frame by frame to extract the time-series image frame sequence; Calculating a timestamp for each frame of the time-sequential image frame sequence; Obtain vehicle inertial sensor data; Dividing the vehicle inertial sensor data into time periods, performing data matching based on the timestamp of each frame, and extracting the vehicle inertial sensor parameters corresponding to the timestamp; Calculate the relative rotation and displacement between multiple frames based on the vehicle inertial sensor parameters; Perform pixel-level motion optical flow estimation on the temporal image frame sequence to obtain pixel-level motion optical flow information; Motion visual jitter elimination and compensation are performed on a temporal image frame sequence based on pixel-level motion optical flow information, the relative rotation and displacement, so as to construct a visual jitter optimized real-time video stream.

4. The data visualization method of the electric vehicle intelligent driving recorder according to claim 1 is characterized in that: The specific steps of step S2 are: Intelligently identify road boundary elements based on visual jitter optimization of real-time video streams and extract road boundary element information; Calculate slope changes and road curvature based on visual jitter optimization of real-time video streams to obtain road condition information; Based on the multi-level scene target information sequence, the road boundary element information, slope changes and road curvature are mined in real time to construct the driving scene road topology map; Optimize real-time video streams based on visual jitter to track dynamic targets and road topology, and extract real-time changing trajectories of dynamic targets and road topology; Based on the real-time change trajectory, multi-dimensional scene topology trend evolution fitting is performed to construct the scene motion trend matrix; Based on the scene motion trend matrix, the driving scene road topology map is rendered in spatiotemporal connection and the scene evolution is smoothed to construct a dynamic driving scene topology rendering model.

5. The data visualization method of the electric vehicle intelligent driving recorder according to claim 1 is characterized in that: The specific steps of step S3 are: Obtain vehicle operating status parameters; calculate real-time vehicle speed, engine power output, vehicle battery status, and motor temperature based on the vehicle operating status parameters to obtain multiple vehicle operating indicators; Conduct multi-parameter correlation analysis on multiple vehicle operation indicators and extract correlation patterns of multiple indicators; Identify real-time driving behavior data based on vehicle operating status parameters; Perform vehicle transient response analysis on multiple vehicle operating indicators based on real-time driving behavior data to obtain real-time vehicle response data; Based on the correlation rules of multiple indicators, a comprehensive status correlation evaluation is performed on the vehicle status response data to build a real-time vehicle status evaluation report.

6. The data visualization method of the electric vehicle intelligent driving recorder according to claim 1, characterized in that: The specific steps of step S4 are: Identify road speed limit signs and regional speed limit regulations based on a dynamic driving scene topology rendering model to obtain road time limit information; Real-time dynamic target recognition is performed based on the dynamic driving scene topology rendering model, and road traffic density is calculated to obtain road traffic density characteristics; Carry out vehicle traffic distribution mining based on road traffic density characteristics and generate vehicle traffic density distribution characteristics; Identify the distribution of road obstacles and changes in road curvature based on the dynamic driving scene topology rendering model, and generate road complexity data; Conduct road safety assessment based on vehicle traffic density distribution characteristics and road complexity data to obtain road safety assessment results; Based on the real-time vehicle status assessment report, multi-objective constraint calculations are performed on the road safety assessment results, and the optimal vehicle speed is predicted to obtain the optimal vehicle speed.

7. The data visualization method of the electric vehicle intelligent driving recorder according to claim 1 is characterized in that: The specific steps of step S5 are: Accurately segment dynamic targets in the topological rendering model of dynamic driving scenes and extract each dynamic target; Calculate the velocity vector, acceleration change rate and direction turning frequency of each dynamic target to obtain the kinematic parameters of each target; Perform intention reasoning and behavior prediction based on the kinematic parameters to generate a set of behavior prediction trajectories for each target point; Multi-path probability calculation is performed on the behavior prediction trajectory set, and credibility weight analysis is performed to extract the maximum probability behavior prediction trajectory of each target.

8. The data visualization method of the electric vehicle intelligent driving recorder according to claim 1, characterized in that: The specific steps of step S6 are: Based on the dynamic driving scene topology rendering model, real-time road topology state evolution recognition is performed to extract dynamic road topology state evolution features; Performing a road driving safety risk analysis based on the dynamic road topology state evolution characteristics and the maximum probability behavior prediction trajectory to generate a road driving risk level; Calculate the optimal safety distance for each target based on the road driving risk level and the optimal vehicle speed to obtain the optimal safety distance for each target; Performing real-time dynamic mapping on the dynamic driving scene topology rendering model according to the optimal safety distance to construct a holographic driving scene model; Perform visual rendering of the vehicle's central control screen based on the holographic driving scene model to perform visualization operations.

9. A data visualization system for an electric vehicle intelligent driving recorder, characterized in that: The method for performing data visualization of the electric vehicle intelligent driving recorder according to claim 1 comprises: The video processing module is used to obtain real-time video streams of driving scenes based on intelligent driving recorders, perform motion visual jitter elimination and compensation, and perceive multi-level scene target features to construct a multi-level scene target information sequence; The scene topology rendering module is used to perform spatiotemporal connection rendering based on multi-level scene target information sequences and smooth scene evolution, thereby constructing a dynamic driving scene topology rendering model; The vehicle status assessment module is used to obtain vehicle operating status parameters, perform multi-parameter correlation analysis and vehicle transient response analysis, and construct a real-time vehicle status assessment report; A multi-objective constraint module is used to perform multi-objective constraint calculations based on a dynamic driving scenario topology rendering model and to predict the optimal vehicle speed based on a real-time vehicle status assessment report to obtain the optimal vehicle speed. The behavior trajectory prediction module is used to calculate the multi-path probability of dynamic targets in the dynamic driving scene topology rendering model, perform credibility weight analysis, and extract the maximum probability behavior prediction trajectory for each target; The visualization rendering module is used to calculate the optimal safety distance for each target based on the maximum probability behavior prediction trajectory and the optimal vehicle speed, and perform visualization rendering on the vehicle's central control screen to perform visualization operations.

Images