Dynamic picture projection system and method based on automobile rearview mirror

Abstract

The invention relates to the technical field of image projection, in particular to a dynamic image projection system and method based on an automobile rearview mirror. The method comprises the following steps: acquiring an automobile rear-view panoramic multispectral image stream, and performing adaptive ambient light interference optimization to generate a multi-level ambient light interference optimized image stream; performing dynamic light intensity change adaptation calculation and nonlinear gamma correction on the multi-level ambient light interference optimization image flow to generate a dynamic light adaptation image frame sequence; performing reverse dynamic parallax prediction on the dynamic light adaptive image frame sequence, performing inter-frame pixel migration reconstruction, and constructing a coherent visual dynamic image sequence; mirror surface curvature information of the automobile rearview mirror is calculated, mirror surface deformation correction and pixel relocation mapping are carried out on the coherent visual dynamic image sequence, and a distortion correction dynamic image sequence is constructed. Clear and coherent real-time rearview mirror projection is realized, the visualization degree of a dangerous vehicle is enhanced, and the driving safety is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image projection, and in particular to a motion picture projection system and method based on a car rearview mirror. Background Art

[0002] With the rapid development of smart cars and assisted driving technologies, traditional rearview mirrors are facing new technological changes. As a crucial component for vehicle safety, rearview mirrors primarily allow drivers to observe road conditions behind and to the sides of the vehicle. However, traditional optical mirrors are limited by external environmental factors such as field of view, weather conditions, and light intensity, making it difficult to provide comprehensive and intuitive visual information in complex driving scenarios. This is particularly true during critical moments such as nighttime driving, inclement weather, or when monitoring blind spots. Traditional rearview mirrors can suffer from blind spots and image blur, seriously compromising driving safety.

[0003] In recent years, with advances in in-vehicle display technology, computer vision, and human-computer interface (HMI) technology, visual assistance systems based on digital projection and image enhancement have been gradually adopted in automotive rearview systems. Motion image projection technology, in particular, projects real-time processed dynamic images directly into the driver's field of view, effectively integrating information collected by the vehicle's perception system to visually reconstruct and dynamically enhance the rearview environment. This motion image-based projection method not only overcomes the physical limitations of traditional rearview mirrors but also provides a more flexible, clear, and intelligent display of rearward images, significantly enhancing the driving experience and driving safety.

[0004] However, most current image-based rearview mirror enhancement systems remain at the static image or video playback level, lacking the ability to quickly respond to dynamic changes in the driving environment. This makes it difficult to efficiently identify and present dynamic events in complex traffic scenarios in real time. Existing motion image projection technology also faces numerous technical challenges, including image stability, projection area calibration, and driver distraction control. Images can be easily blurred at high speeds or when the vehicle is vibrating. Inconsistencies between the projected content and the driver's perspective can interfere with driving judgment. Image overload can also lead to information fatigue. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention proposes a dynamic image projection system and method based on the use of a car rearview mirror to solve at least one of the above technical problems.

[0006] To achieve the above object, the present invention provides a method for projecting a moving image based on a car rearview mirror, comprising the following steps: Step S1: Obtaining a vehicle rearview panoramic multispectral image stream and performing adaptive ambient light interference optimization to generate a multi-level ambient light interference optimized image stream; Step S2: performing dynamic light intensity change adaptation calculation and nonlinear gamma correction on the multi-level ambient light interference optimized image stream to generate a dynamic light adaptation image frame sequence; Step S3: performing reverse dynamic disparity prediction on the dynamic light adaptation image frame sequence and performing inter-frame pixel migration reconstruction to construct a coherent visual animation sequence; Step S4: Calculate the mirror curvature information of the rearview mirror of the car, and perform mirror deformation correction and pixel relocation mapping on the coherent visual animation sequence to construct a distortion-corrected animation sequence; Step S5: Calculate the position of the moving object in multiple time periods and predict its behavior trajectory based on the distortion-corrected moving image sequence to generate a predicted behavior trajectory of the moving object; Step S6: Calculate the collision risk based on the predicted trajectory of the moving object, mark high-risk target objects, and highlight and visualize the distortion-corrected animated image sequence.

[0007] This invention achieves all-weather rearview sensing capabilities through multispectral image acquisition (which may include visible light, infrared, and low-light channels), addressing the failure of traditional cameras in harsh scenarios such as nighttime, rain, fog, and backlighting. An adaptive ambient light interference optimization mechanism dynamically adjusts image brightness and contrast, suppressing strong light reflections and interference from rear lights, improving the clarity and perceptual stability of the original image stream. A multi-level image optimization structure provides a robust foundation for subsequent image analysis modules, ensuring that key targets are not obscured or misidentified due to light fluctuations. A dynamic light intensity adaptation algorithm adjusts the image brightness response in real time across different scenarios, enabling the system to adapt to extreme brightness changes such as sudden entry into tunnels and areas of intense light. Gamma nonlinear correction enhances the representation of image details across different brightness ranges, preventing loss of dark image information in low-light environments. The output dynamic light adaptation image frames provide a unified light perception baseline for subsequent animation construction, preventing flickering or drifting during brightness changes. Inverse dynamic parallax prediction accurately estimates the depth variation trend between static backgrounds and dynamic targets in an image. Pixel migration reconstruction technology breaks the traditional frame-to-frame dependency, constructing a smooth visual animation sequence and eliminating jitter and frame skipping. The resulting animation features smooth visuals and a sense of three-dimensional depth, enhancing the driver's intuitive ability to react to environmental changes in the mirror. The curved design of rearview mirrors can easily cause image edge distortion and visual misjudgment. This step performs a geometric inverse transformation based on the actual mirror parameters to restore the true spatial proportions. Pixel repositioning ensures that objects in the animation correspond to their actual positions in the mirror projection, significantly reducing delays in hazard assessment caused by projection errors. The corrected image exhibits excellent spatial consistency, eliminating the need for drivers to adapt to the "misalignment of virtual and real" issues and improving visual trust. Multi-period position recognition combined with time-window dynamic modeling can more accurately identify the movement patterns of approaching small targets such as vehicles, bicycles, and electric scooters. The trajectory prediction function generates a "predicted path" based on historical position and speed trends, providing drivers with a "one-second-in-the-future" decision-making reference. Early warning of approaching objects can be used to avoid dangerous situations such as sudden appearances from the side and rear, and is a key component of achieving "enhanced perception" in rearview mirrors. A collision risk calculation model comprehensively evaluates target speed, acceleration, relative distance, and predicted path to determine whether it intersects with the vehicle's trajectory. High-risk objects are visually highlighted (e.g., dynamic outlines, color changes, pulsing haloes), allowing drivers to receive hazard information without focusing their attention.

[0008] In this specification, a motion picture projection system based on a car rearview mirror is provided, which is used to execute the motion picture projection method based on a car rearview mirror as described above, including: Light interference optimization module, used to obtain the car rear view panoramic multispectral image stream, and perform adaptive ambient light interference optimization to generate a multi-level ambient light interference optimized image stream; An image processing module is used to perform dynamic light intensity change adaptation calculation and nonlinear gamma correction on the multi-level ambient light interference optimized image stream to generate a dynamic light adaptation image frame sequence; The inter-frame pixel reconstruction module is used to perform reverse dynamic disparity prediction on the dynamic light adaptation image frame sequence and perform inter-frame pixel migration reconstruction to construct a coherent visual animation sequence; The distortion correction module is used to calculate the mirror curvature information of the car's rearview mirror, and perform mirror deformation correction and pixel relocation mapping on a coherent visual animation sequence to construct a distortion-corrected animation sequence; The trajectory prediction module is used to calculate the multi-time position of the moving object and predict the behavior trajectory based on the distortion-corrected dynamic image sequence to generate the predicted behavior trajectory of the moving object; The highlight visualization module is used to calculate collision risk based on the predicted trajectory of moving objects, mark high-risk target objects, and highlight and visualize the distortion-corrected animation sequence.

[0009] This invention utilizes an ambient light interference optimization mechanism to identify strong light, reflections, and high-brightness background interference (such as high beams and direct sunlight) in real time, dynamically masking and balancing it. Its multi-layered image output provides a multi-dimensional reference image source for subsequent image processing, improving the system's robustness to interference and ensuring image stability and continuity. A dynamic adaptation algorithm aligns local brightness and overall illumination across consecutive image frames, ensuring brightness consistency between images and avoiding visual jumps. Gamma correction enhances dark detail and highlight suppression, improving image contrast and texture clarity. This provides a unified, stable, and faithfully reproduced image sequence, providing an ideal input source for animated image construction and enhancing rearview mirror visual clarity and responsiveness. A disparity prediction algorithm identifies relative depth changes between moving objects and backgrounds within image frames, enabling accurate image temporal vector modeling. Pixel migration technology generates fitted frames from non-consecutive frames, effectively eliminating animation artifacts such as stuttering, ghosting, and frame skipping. The resulting visual animation is coherent and natural, creating an augmented reality-like rearview animation experience and improving the driver's spatial judgment and attention acquisition. The system accurately captures the mirror's actual surface parameters and combines them with a dynamic image projection deformation model to achieve precise image geometry correction. A pixel repositioning algorithm ensures that the projected image closely corresponds to the actual physical location, preventing object position misjudgment due to distortion. The corrected dynamic image projection eliminates parallax errors without changing traditional rearview mirror usage, achieving a hybrid visual experience combining "real location + information enhancement." Multi-period behavioral modeling is performed based on the target's position sequence, speed variation, and trajectory curvature within the dynamic image, enabling high-precision prediction of dynamic objects. The system accommodates a wide range of targets, including vehicles, bicycles, pedestrians, small animals, and other dynamic entities of varying sizes. Predicted trajectories serve as a prerequisite for behavioral risk assessment, laying the foundation for proactive system responses (such as target highlighting and voice warnings). Collision risk calculation analyzes potential collision trends based on relative motion vectors, angular proximity, and speed differences. High-risk targets are visually highlighted in the dynamic image through color enhancement, outlines, and dynamic icons, creating a head-up display (HUD)-like visual prompt. This effectively directs the driver's attention to potential hazards, improving driving safety, particularly in complex urban traffic conditions or high-speed lane changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 This is a schematic flow chart of the steps of a method for projecting a moving image based on a car rearview mirror 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

[0011] 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.

[0012] This application provides a system and method for projecting an image based on a car rearview mirror. The system and method include, but are not limited to, the following: mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc. that are equipped with the system, 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.

[0013] See also Figures 1 to 4 The present invention provides a method for projecting a moving picture based on a car rearview mirror, the method comprising the following steps: Step S1: Obtaining a vehicle rearview panoramic multispectral image stream and performing adaptive ambient light interference optimization to generate a multi-level ambient light interference optimized image stream; Step S2: performing dynamic light intensity change adaptation calculation and nonlinear gamma correction on the multi-level ambient light interference optimized image stream to generate a dynamic light adaptation image frame sequence; Step S3: performing reverse dynamic disparity prediction on the dynamic light adaptation image frame sequence and performing inter-frame pixel migration reconstruction to construct a coherent visual animation sequence; Step S4: Calculate the mirror curvature information of the rearview mirror of the car, and perform mirror deformation correction and pixel relocation mapping on the coherent visual animation sequence to construct a distortion-corrected animation sequence; Step S5: Calculate the position of the moving object in multiple time periods and predict its behavior trajectory based on the distortion-corrected moving image sequence to generate a predicted behavior trajectory of the moving object; Step S6: Calculate the collision risk based on the predicted trajectory of the moving object, mark high-risk target objects, and highlight and visualize the distortion-corrected animated image sequence.

[0014] This invention achieves all-weather rearview sensing capabilities through multispectral image acquisition (which may include visible light, infrared, and low-light channels), addressing the failure of traditional cameras in harsh scenarios such as nighttime, rain, fog, and backlighting. An adaptive ambient light interference optimization mechanism dynamically adjusts image brightness and contrast, suppressing strong light reflections and interference from rear lights, improving the clarity and perceptual stability of the original image stream. A multi-level image optimization structure provides a robust foundation for subsequent image analysis modules, ensuring that key targets are not obscured or misidentified due to light fluctuations. A dynamic light intensity adaptation algorithm adjusts the image brightness response in real time across different scenarios, enabling the system to adapt to extreme brightness changes such as sudden entry into tunnels and areas of intense light. Gamma nonlinear correction enhances the representation of image details across different brightness ranges, preventing loss of dark image information in low-light environments. The output dynamic light adaptation image frames provide a unified light perception baseline for subsequent animation construction, preventing flickering or drifting during brightness changes. Inverse dynamic parallax prediction accurately estimates the depth variation trend between static backgrounds and dynamic targets in an image. Pixel migration reconstruction technology breaks the traditional frame-to-frame dependency, constructing a smooth visual animation sequence and eliminating jitter and frame skipping. The resulting animation features smooth visuals and a sense of three-dimensional depth, enhancing the driver's intuitive ability to react to environmental changes in the mirror. The curved design of rearview mirrors can easily cause image edge distortion and visual misjudgment. This step performs a geometric inverse transformation based on the actual mirror parameters to restore the true spatial proportions. Pixel repositioning ensures that objects in the animation correspond to their actual positions in the mirror projection, significantly reducing delays in hazard assessment caused by projection errors. The corrected image exhibits excellent spatial consistency, eliminating the need for drivers to adapt to the "misalignment of virtual and real" issues and improving visual trust. Multi-period position recognition combined with time-window dynamic modeling can more accurately identify the movement patterns of approaching small targets such as vehicles, bicycles, and electric scooters. The trajectory prediction function generates a "predicted path" based on historical position and speed trends, providing drivers with a "one-second-in-the-future" decision-making reference. Early warning of approaching objects can be used to avoid dangerous situations such as sudden appearances from the side and rear, and is a key component of achieving "enhanced perception" in rearview mirrors. A collision risk calculation model comprehensively evaluates target speed, acceleration, relative distance, and predicted path to determine whether it intersects with the vehicle's trajectory. High-risk objects are visually highlighted (e.g., dynamic outlines, color changes, pulsing haloes), allowing drivers to receive hazard information without focusing their attention.

[0015] In the embodiment of the present invention, see Figure 1 , is a schematic flow chart of the steps of a method for projecting a moving image based on a car rearview mirror according to the present invention. In this example, the steps of the method for projecting a moving image based on a car rearview mirror include: Step S1: Obtaining a vehicle rearview panoramic multispectral image stream and performing adaptive ambient light interference optimization to generate a multi-level ambient light interference optimized image stream; In this embodiment, a multispectral camera is used to capture real-time images while the vehicle is in motion. Ensure the acquisition frequency is high enough to capture dynamic scene changes; a recommended data acquisition frequency of 30 frames per second (FPS) is recommended. Multispectral image streams are continuously captured in various driving environments (e.g., daytime, cloudy, and nighttime), and the timestamp and environmental conditions of each frame are recorded for subsequent analysis. Before performing ambient light interference optimization, the captured image stream must first be analyzed for ambient light characteristics. This involves identifying image features such as color temperature, light intensity, and light dispersion. These features can be extracted using image processing algorithms. Image histogram analysis is used to calculate the average brightness and color distribution of each frame to determine the current ambient light characteristics. An appropriate image optimization algorithm is selected for adaptive ambient light interference optimization. Common algorithms include adaptive histogram equalization (CLAHE) and gamma correction, which effectively enhance image contrast and brightness. If the analysis results indicate a high ambient light color temperature (e.g., 6000K), the CLAHE algorithm can be used to enhance local image contrast to overcome interference caused by uneven illumination. Apply the selected adaptive optimization algorithm to each frame to ensure that each frame is optimized to improve overall image quality. The optimization process should take into account different ambient light characteristics and dynamically adjust the optimization parameters.

[0016] Step S2: performing dynamic light intensity change adaptation calculation and nonlinear gamma correction on the multi-level ambient light interference optimized image stream to generate a dynamic light adaptation image frame sequence; In this embodiment, light intensity variation characteristics are analyzed for a multi-level ambient light interference-optimized image stream. This process involves calculating the average brightness value for each frame to identify intensity variation trends. This can be achieved using an image processing library such as OpenCV. For each frame, the average brightness value is calculated. By traversing the multi-level image stream, the average brightness value of each frame is recorded frame by frame, and the intensity variation between adjacent frames is calculated. The magnitude and direction of the intensity variation are recorded to prepare for dynamic intensity adaptation. A dynamic intensity adaptation strategy is developed based on the intensity variation characteristics. Linear interpolation or weighted averaging can be used to adjust the intensity of the current frame to be consistent with the previous frame to ensure smooth intensity variation. After dynamic intensity adaptation, nonlinear gamma correction is performed to enhance image contrast and detail. The gamma value is dynamically adjusted based on the intensity adaptation value. For images with lower intensity, the gamma value can be appropriately increased to improve brightness; for images with higher intensity, the gamma value can be reduced to prevent overexposure. The calculated gamma value is then applied to each frame for nonlinear correction. To ensure that the correction process is stable and can maintain image quality under different lighting conditions, the gamma correction formula is applied pixel by pixel to each frame of the image, calculating the value of each pixel in the output image to generate the corrected image frame.

[0017] Step S3: performing reverse dynamic disparity prediction on the dynamic light adaptation image frame sequence and performing inter-frame pixel migration reconstruction to construct a coherent visual animation sequence; In this embodiment, a method for reverse dynamic disparity prediction is defined. Disparity refers to the change in the position of an object on the image plane when the same object is observed at two different time points. The prediction model can be calculated based on historical disparity data using a time series analysis method. A simple linear regression model is used to predict the disparity value of the next frame based on the disparity change rate of the previous frames. The input of the model is set to the disparity value of the previous n frames to generate predictions for the next n frames. Before starting the prediction, the existing disparity value needs to be calculated. For a dynamic light adaptation image frame sequence, the disparity between the current frame and the previous frame is calculated by comparing the pixel positions between adjacent frames. Assuming the image resolution is 1920x1080 pixels, if the position of an object in the first frame is (x1, y1) and in the second frame is (x2, y2), the disparity is calculated as: D=(x1-x2, y1-y2), and dynamic prediction is performed using historical disparity data. A simple linear regression model is used to predict the disparity value of the next frame. Assuming the disparity values of the previous two frames are D1 and D2, the disparity of the next frame, D3, is predicted as: D3 = D2 + (D2 - D1). Based on the predicted disparity values, the dynamic light adaptation image frame sequence is reconstructed by inter-frame pixel transfer. The goal of this transfer is to adjust the pixel positions in the current frame to the predicted disparity positions to achieve reconstruction. The transfer formula is: I new (x, y) = I old (x + Dx, y + Dy), where Dx and Dy are offsets calculated based on the predicted disparity. For each frame in the dynamic light adaptation image frame sequence, pixel transfer is performed using the calculated disparity value. For each pixel, a new position is calculated based on its position in the current frame and the predicted disparity value. The reconstructed image frames are recorded to generate a coherent visual animation sequence. This ensures that the transferred reconstruction of each frame maintains image coherence and presents a smooth visual effect.

[0018] Step S4: Calculate the mirror curvature information of the rearview mirror of the car, and perform mirror deformation correction and pixel relocation mapping on the coherent visual animation sequence to construct a distortion-corrected animation sequence; In this embodiment, the geometric shape of a vehicle's rearview mirror is determined, and an appropriate method is selected to measure its mirror curvature. Generally speaking, rearview mirrors are curved surfaces, and their precise three-dimensional shape data can be obtained using optical measurement tools (such as laser scanners). A high-precision laser scanner is used to scan the rearview mirror to obtain point cloud data of its surface. During scanning, the data sampling frequency is set to 1000Hz to ensure sufficient point cloud information is obtained to cover the entire mirror surface. Based on the obtained point cloud data, the mirror curvature information is calculated using a surface fitting algorithm (such as the least squares method). By analyzing the neighborhood of each point in the point cloud, the normal vector is calculated and the curvature value is derived. Specifically, samples are taken at the center and edge of the mirror surface. Assuming the curvature value at the center is 500mm and at the edge is 700mm, the curvature information at different locations is recorded for subsequent correction. Based on the calculated mirror curvature information, a mirror distortion correction method is defined. The goal of mirror distortion correction is to restore the image after mirror reflection to a realistic visual effect. This is typically achieved through an inverse mapping algorithm. An inverse transformation model is established to adjust the coordinates of each pixel based on the curvature of the mirror surface, ensuring that the final image accurately reflects the object's true position. A continuous visual animation sequence is processed frame by frame, and a mirror distortion correction algorithm is applied. For each frame, the new position of each pixel is calculated and repositioned based on the mirror curvature.

[0019] Step S5: Calculate the position of the moving object in multiple time periods and predict its behavior trajectory based on the distortion-corrected moving image sequence to generate a predicted behavior trajectory of the moving object; In this embodiment, an object detection algorithm (such as YOLO, SSD, or Faster R-CNN) is used to identify and track moving objects in images. Ensure that the model can work effectively under different lighting conditions. The YOLOv5 model is selected for training, using a dataset containing multiple target categories such as vehicles and pedestrians so that the model can accurately identify moving objects in different scenarios. The distortion-corrected animated image sequence is processed frame by frame, and the object detection model is called to detect and record the position of the moving object in each frame. The recorded position should include the object's bounding box coordinates, its category, and timestamp information. Based on the extracted moving object position data, a behavior trajectory prediction model is defined. Methods based on time series analysis, such as long short-term memory networks (LSTMs), can be used to capture the object's motion patterns and behavioral characteristics. The LSTM model's input is set to the object position data of the previous n frames to generate predicted trajectories for the next n frames. When training the model, a historical dataset, including labeled object position and behavior data, is required so that the model can effectively learn behavioral patterns under different motion states. The dataset used for model training should include motion trajectories from various scenarios. Training parameters should be set, such as a learning rate of 0.001, a batch size of 32, and 50 training cycles. Historical trajectory data should be fed into the trained LSTM model to generate predicted trajectories for future moving objects. The predicted results should be multi-time period locations to facilitate subsequent visualization and behavioral analysis.

[0020] Step S6: Calculate the collision risk based on the predicted trajectory of the moving object, mark high-risk target objects, and highlight and visualize the distortion-corrected animated image sequence.

[0021] In this embodiment, the minimum distance between the predicted trajectory of each moving object and the vehicle is calculated. This can be achieved by calculating the Euclidean distance between the object's expected position and the vehicle's current position. Based on the calculated distance and warning time, the collision risk level of each moving object is assessed. If R exceeds a set risk threshold (e.g., 2.0), the object is marked as a high-risk target. The collision risk levels of all moving objects are traversed, and high-risk targets are marked according to the set threshold. The information of high-risk target objects is structured for subsequent processing and visualization. If the risk level of object 1 is 2.5, which exceeds the threshold, it is marked as "high risk" and its bounding box coordinates and other related information are recorded. The marked high-risk target objects are recorded and a high-risk target list is generated for subsequent visualization. The recorded high-risk targets are [(object 1, high risk), (object 2, low risk)] so that they can be highlighted in subsequent steps. An appropriate visualization processing method is selected to highlight the high-risk target objects. Image overlay technology can be used to highlight high-risk targets in the animation sequence. High-risk targets are displayed in red to make them easily visible in the image, while maintaining a moderate degree of transparency to avoid obscuring background information. Each frame of the distortion-corrected animation sequence is processed, iterating over the high-risk targets and applying a highlighting effect. A highlight mark is added to the border of each target.

[0022] 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: Based on the low-light multispectral camera unit, high dynamic range real-time sampling is performed to obtain the vehicle rear view panoramic multispectral image stream; Performing multi-level light characteristic analysis on the vehicle rearview panoramic multispectral image stream to identify the ambient light color temperature, light intensity, and dispersion; Performing ambient light scene analysis based on the ambient light color temperature, light intensity, and dispersion to identify the current ambient light scene; Adaptive ambient light interference optimization is performed on the vehicle rearview panoramic multispectral image stream according to the ambient light scene to generate a multi-level ambient light interference optimized image stream.

[0023] In this embodiment, real-time sampling is performed in an actual vehicle rearview environment to acquire a panoramic multispectral image stream. This process takes into account varying ambient light conditions (such as daylight, cloudy skies, and nighttime) to ensure that the camera unit's dynamic range can cover these variations. Under different lighting conditions (such as direct sunlight, shadows, and nighttime streetlights), 30 seconds of image streams are captured in each case to obtain sufficient data for subsequent analysis. The multispectral image streams obtained through real-time sampling are stored on high-performance storage media to ensure data integrity and fast access. Furthermore, the data format is standardized (such as TIFF or RAW format) for subsequent processing. SSD storage devices are used to support high data transfer rates and ensure that no frames are dropped during image stream acquisition. Light characteristic parameters to be analyzed are defined, including ambient light color temperature, light intensity, and light dispersion. These parameters are used in subsequent ambient light scene analysis to help identify the current lighting conditions. The color temperature range is set to 1000K to 10000K, light intensity is expressed in lux, and dispersion is measured by the width of the spectral distribution. Use image processing algorithms (such as image segmentation, filtering, and feature extraction) to analyze the multispectral image stream and extract information about the color temperature, intensity, and dispersion of the ambient light. This can be achieved by performing spectral analysis on the spectral data in the image. Principal component analysis (PCA) is used to process the spectral data to extract key light features and calculate the color temperature and intensity of the ambient light. For example, assume that in a given image, the calculated color temperature is 5500K and the light intensity is 3000 Lux. The extracted light characteristic parameters are recorded and an analysis report is generated describing the current ambient light characteristics and their changes. This data provides the basis for subsequent ambient light scene analysis. Changes in light intensity over different time periods are recorded, and color temperature fluctuations in specific environments are marked for subsequent analysis. Based on the extracted ambient light color temperature, intensity, and dispersion data, an ambient light scene model is constructed. This model should be able to reflect environmental characteristics under different lighting conditions and identify possible light scene types (such as sunny, cloudy, and nighttime). Set thresholds for different color temperatures and light intensities, defining sunny color temperatures as 5000K to 6500K, cloudy color temperatures as 6500K to 8000K, and nighttime color temperatures below 3000K. Use classification algorithms (such as KNN and decision trees) to analyze the ambient light scene and identify the current ambient light scenario. Input the extracted light characteristic parameters into the model for scene classification. If the analysis results show the current color temperature is 5500K and the light intensity is 3000 Lux, the model identifies the scene as "sunny." Based on the identified ambient light scenario, select an appropriate image optimization algorithm to achieve adaptive ambient light interference reduction. These algorithms should be able to adjust image parameters according to different scenarios to improve image quality. Selecting the adaptive histogram equalization (CLAHE) algorithm can effectively improve image contrast, especially in uneven lighting conditions. Optimization parameters such as brightness adjustment, contrast enhancement, and color correction are set for different ambient light scenarios.These parameters are dynamically adjusted based on the characteristics of the ambient light. Under high-intensity, sunny conditions, the brightness is reduced to 0.2, while the contrast is increased to 1.5 to prevent overexposure. The vehicle's rearview panoramic multispectral image stream is processed and an adaptive ambient light interference optimization algorithm is applied to generate a multi-level, optimized image stream. This ensures that each image frame is optimized to improve overall image quality. The processed image stream displays improved detail and color reproduction under sunny conditions, reducing interference caused by strong light.

[0024] In this embodiment, the steps of adaptively optimizing the ambient light interference of the vehicle rearview panoramic multispectral image stream according to the ambient light scene to generate a multi-level ambient light interference optimized image stream are as follows: Performing illumination interference blur calculation on the ambient light scene to identify the interference blur degree of the current scene; performing adaptive light compensation analysis according to the interference blur degree to obtain light interference compensation parameters; Identify the ambient light interference area of the car rearview panoramic multispectral image stream and extract the sub-image of the light interference area; Adaptively suppress and optimize the sub-image of the light interference area according to the light interference compensation parameter to generate a light interference suppression optimized sub-image; The vehicle rearview panoramic multispectral image stream is globally fused and reconstructed based on the light interference suppression optimized sub-image to generate a multi-level ambient light interference optimized image stream.

[0025] In this embodiment, the degree of blur can be quantified using metrics such as image contrast, sharpness, and clarity. Common metrics include Laplace transform and edge detection algorithms. A blur threshold is set, and the image variance is calculated using the Laplace operator. A smaller variance indicates a more blurred image. The acquired multispectral image stream is processed, and an edge detection algorithm (such as Canny edge detection) is applied to extract edge information and calculate the degree of blur. Blur calculation should be performed on images at multiple time points to obtain reliable results. For example, if the variance obtained after Laplace transform in a certain image is 15, it indicates a high degree of blur, which may affect subsequent visual effects. Based on the identified degree of blur caused by illumination interference, adaptive light compensation parameters are set. These parameters should be able to compensate for image blur caused by illumination interference and restore image clarity. The compensation parameters are set as brightness adjustment coefficient, contrast enhancement coefficient, and sharpness enhancement coefficient. Image enhancement algorithms (such as histogram equalization and gamma correction) are used to perform adaptive light compensation analysis. By quantifying the degree of blur, the image light compensation parameters are dynamically adjusted. If the blurriness is 30, set the brightness adjustment factor to 1.2 and the contrast enhancement factor to 1.5 to enhance the visual quality of the image. The calculated light compensation parameters should be recorded, and a report should be generated describing the source of the compensation parameters and their impact on image quality. This will provide a basis for subsequent identification of light interference areas. For a blurriness of 30, the final calculated compensation parameters are recorded as brightness 1.2, contrast 1.5, and sharpness 1.3. Based on the generated light compensation parameters, identify areas of ambient light interference in the image. Use a threshold-based method or region growing algorithm to identify light interference areas. Set a threshold; if the blurriness of an area exceeds the set value, mark it as a light interference area. Extract the identified light interference area and generate a sub-image of the light interference area. Ensure that the extracted area contains sufficient information for subsequent optimization. If the interference area is identified as part of the lower left corner of an image, extract this area and save it as a separate sub-image. Adaptively suppress and optimize the light interference area sub-image based on the light interference compensation parameters. Select an appropriate image processing algorithm (such as non-local mean denoising, bilateral filtering, etc.) for optimization. Select the non-local mean denoising algorithm to effectively reduce blur caused by light interference. Set the optimization algorithm parameters, including denoising intensity and window size. Ensure that these parameters are dynamically adjusted based on the light compensation parameters to achieve the best results. Set the denoising intensity to 0.5 and the window size to 7 to ensure that both detail and denoising effects are taken into account during processing. Process the extracted sub-image of the light interference area and apply the adaptive suppression optimization algorithm to generate an optimized light interference suppression sub-image. Ensure that the optimized image can effectively improve clarity. The processed sub-image shows a significant improvement in clarity, with the blur reduced to 10. Record this result for subsequent analysis.Based on the optimized sub-images for light interference suppression, the entire vehicle rearview panoramic multispectral image stream is globally fused and reconstructed. An appropriate image fusion algorithm (such as weighted averaging or Laplacian pyramid) is selected for processing. The weighted averaging method is selected to fuse each image based on its clarity and other features for optimal visual quality. The parameters of the fusion algorithm are set, including weight assignment and fusion strategy. These parameters are dynamically adjusted based on the characteristics of the optimized sub-images. Weights are set based on clarity, with images with lower blurriness having a weight of 0.7 and images with higher blurriness having a weight of 0.3. A global fusion algorithm is executed, combining the optimized sub-images for light interference suppression with the original image stream to generate a multi-layered image stream optimized for ambient light interference. This ensures that the final output image stream has good visual quality and clarity. The resulting optimized image stream displays improved detail and contrast under different lighting conditions, improving overall image quality.

[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: Perform global frame-by-frame decomposition of the multi-level ambient light interference optimized image stream to generate a time-series frame image sequence; Calculating the average brightness value of each frame image in the time-sequential frame image sequence; Calculating the brightness difference between frames based on the average brightness value to generate brightness difference information between frames; Performing an ambient light intensity fluctuation trend analysis on the current ambient light blur scene to extract ambient light intensity fluctuation trend characteristics; Based on the characteristics of the ambient light intensity fluctuation trend, the inter-frame brightness difference signal is dynamically adapted to calculate the brightness change to obtain the inter-frame brightness video compensation value; The multi-level ambient light interference optimized image stream is subjected to nonlinear gamma correction according to the inter-frame brightness video compensation value to generate a dynamic light adaptation image frame sequence.

[0027] In this embodiment, each frame is extracted from a multi-level ambient light interference-optimized image stream. The goal of global frame-by-frame decomposition is to convert a continuous image stream into a sequence of independent frame images for subsequent analysis and calculation. Assuming the image stream contains 100 frames, image processing software or a custom algorithm is used to extract and store each frame, ensuring that each frame can be accessed and analyzed individually. During the frame-by-frame decomposition process, ensure that all frames maintain a consistent format. The images can be uniformly converted to RGB or grayscale format to facilitate subsequent brightness calculation and analysis. Using grayscale format for processing reduces computational complexity, improves efficiency, and ensures the accuracy of subsequent processing results. A method for calculating the average brightness value of each frame is determined. Typically, the average brightness value is obtained by summing the brightness values of each pixel and dividing by the total number of pixels. The extracted image sequence is processed frame by frame to calculate the average brightness value of each frame. This process can be efficiently implemented using an image processing library such as OpenCV. Loop through each frame, calculating and recording the average brightness value of each frame. Assume that the average brightness value of the first frame is 120 and the average brightness value of the second frame is 130. Define a method for calculating the brightness difference between frames. The brightness difference between frames can be obtained by calculating the difference in the average brightness values of adjacent frames. The calculated average brightness value array is used for frame-by-frame brightness difference calculations. By traversing the brightness value array, the brightness difference between adjacent frames is calculated and the results are recorded. If the brightness difference between the first and second frames is 10, and the difference between the second and third frames is -5, the difference information is recorded as an array [10, -5, ...]. This brightness difference information is recorded to generate a list containing all frame difference values for subsequent analysis of ambient light intensity fluctuation trends. A method is defined to calculate the characteristics of ambient light intensity fluctuation trends. Fluctuation trends can be identified by calculating the standard deviation, average, and maximum value of the brightness difference between frames. The standard deviation is used as a measure of fluctuation; a large standard deviation indicates significant light intensity fluctuations. A statistical method is used to analyze the brightness difference between frames, calculating features such as the standard deviation, average, and maximum value to extract characteristics of ambient light intensity fluctuation trends. For example, if the calculated standard deviation is 8, the average is 2, and the maximum is 20, this indicates significant ambient light intensity fluctuations and requires attention to their impact on the image. Based on the extracted ambient light intensity fluctuation trend characteristics, the dynamic light intensity change adaptation calculation method is defined. The adaptation value can be calculated using linear regression or weighted average method. The adaptation value calculation formula is set to A=k•ΔL+b, where A is the adaptation value, k is the fluctuation intensity coefficient, and b is the offset. According to the aforementioned fluctuation trend characteristics, dynamic light intensity change adaptation calculation is performed to generate inter-frame brightness video compensation values. Ensure that the calculation process can dynamically reflect light intensity changes. If the fluctuation intensity coefficient k is 1.5 and the offset b is 2, the compensation value obtained by the adaptation calculation is A=1.5•2+2=5. Based on the dynamic light adaptation value, nonlinear gamma correction is performed on the multi-level ambient light interference optimization image stream. Gamma correction can be achieved through , where 𝛾 is the gamma value, which is dynamically adjusted based on the adaptation value. Dynamic gamma correction is applied to each image frame, adjusting the gamma value based on the previously calculated compensation value. This ensures that the corrected image improves both brightness and contrast. Each frame is corrected based on the calculated compensation value. If the current frame compensation value is 5, the gamma value is adjusted to 1.5 for gamma correction. The gamma-corrected image sequence is output to generate a dynamic light adaptation image frame sequence. This ensures that the final image quality meets the visual requirements. The resulting image sequence displays better detail and color reproduction under different lighting conditions, improving overall image quality.

[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: Perform deep visual recognition on the dynamic light adaptation image frame sequence and perform convolutional network semantic segmentation to extract static obstacles and moving objects in the scene; Perform dynamic optical flow tracking and frame-by-frame position calculation on moving objects to obtain multi-frame pixel position points of moving objects; Performing multi-time point disparity calculation based on the multi-frame pixel position points to generate multi-frame object disparity; Calculating a temporal disparity change rate of the multi-frame object disparity; performing reverse dynamic disparity prediction based on the temporal disparity change rate to generate a reverse dynamic disparity difference value; Based on the inverse dynamic parallax difference, the dynamic light adaptation image frame sequence is reconstructed by inter-frame pixel migration to generate a dynamic interpolation reconstructed image; Perform temporal motion graphics fitting based on the dynamic interpolation reconstructed images to construct a coherent visual motion graphics sequence.

[0029] In this embodiment, a suitable convolutional neural network (CNN) model is selected for semantic segmentation. Common models include U-Net, SegNet, or Deep Lab. These models can effectively extract static obstacles and moving objects in images. When training the convolutional network, a well-labeled image dataset is required so that the model can learn the characteristics of both static and dynamic objects. Training is performed using an image dataset containing cars, pedestrians, traffic signs, and other objects to ensure that the model can accurately identify objects in different scenes. A sequence of dynamic light adaptation image frames is input into the trained deep learning model for frame-by-frame semantic segmentation. The model outputs object category labels for each frame, identifying static obstacles (such as walls and streetlights) and moving objects (such as pedestrians and vehicles). For a single image frame, the model may label pedestrians as "moving objects" and buildings as "static obstacles." The segmentation output forms a label map of the same size as the input image. An appropriate optical flow calculation method is selected for tracking dynamic objects. Common optical flow algorithms include the Lucas-Kanade method and the Horn-Schunck method, which calculate the motion vectors of objects in an image. The Lucas-Kanade method is used to calculate the optical flow between the current and previous frames to obtain motion information of moving objects. The dynamic light adaptation image frame sequence is processed frame by frame, calculating the optical flow of the moving object in each frame. The frame-by-frame position of the object is then calculated based on the optical flow information. By tracking the object's motion, its pixel position in the image sequence can be determined. If a pedestrian's position is (x1, y1) in the first frame and (x2, y2) in the second frame, the motion vector calculated through optical flow will help determine the pedestrian's new position in each frame. The pixel positions of the moving objects in each frame are recorded to generate a list containing the positions of all moving objects. This provides the necessary data for subsequent disparity calculations. A method for calculating multi-frame disparity is defined. Disparity is the difference in the position of an object on the image plane when viewed from two different time points or viewpoints. Disparity can be calculated using the pixel positions of an object in two frames. Disparity can be obtained by calculating the difference in pixel positions of the same object in adjacent frames. The formula is D=x frame1-x frame2. The pixel positions of the recorded multiple frames are calculated to generate the corresponding multi-frame object disparity. The disparity between adjacent frames is calculated frame by frame and the disparity value of each frame is recorded. If the position of the pedestrian in the first frame is (100, 200) and in the second frame is (110, 205), then the disparity D is (100-110,200-205)=(-10,-5). Define the calculation method of the parallax change rate. The parallax change rate can be obtained by calculating the difference between adjacent parallax values and normalizing it to reflect the speed of parallax change. Set the calculation formula to R= ,in The disparity value of the i-th frame is processed frame by frame, and t is the time interval. The disparity of objects in multiple frames is processed frame by frame, and the disparity change rate of each frame is calculated. The change rate of each frame is recorded for subsequent analysis. Based on the calculated time-series disparity change rate, a reverse dynamic disparity prediction model is constructed. Simple linear prediction or more complex machine learning models, such as regression analysis, can be used to predict future disparity values. The linear regression model is used to predict the disparity value of the next frame, performing linear interpolation based on the change rates of the previous frames. Based on the predicted value obtained from the change rate calculation, reverse dynamic disparity is calculated to generate reverse dynamic disparity difference values. This process should be able to predict disparity changes for the next few frames. If the current frame disparity is (-8, -4), the next frame disparity is predicted to be (-6, -3) based on the change rate (2, 1). A method for inter-frame pixel migration reconstruction is defined. Based on the reverse dynamic disparity difference values, the pixel positions in the current frame are adjusted and migrated to the predicted target positions to generate a reconstructed image. If a pixel in the current frame is at position (x, y) and the inverse dynamic parallax is (-6, -3), the new pixel position is (x+6, y+3). The dynamic light adaptation image frame sequence is processed frame by frame, and the inverse dynamic parallax difference is applied to pixel migration to generate a dynamic interpolated reconstructed image. For each frame, the pixel positions are adjusted based on the inverse parallax calculation results to generate a new frame image. The generated dynamic interpolated reconstructed images are recorded to generate a time-series image sequence for subsequent time-series motion graphics fitting. The generated reconstructed image sequence is recorded for subsequent visualization and analysis. An appropriate time-series motion graphics fitting method is selected to ensure that the generated image sequence displays smoothly and forms a coherent visual motion image. Techniques such as temporal interpolation and keyframe interpolation can be used. Keyframe interpolation is used to generate smooth transitions between the dynamic interpolated reconstructed images. Time-series motion graphics fitting is performed on the generated dynamic interpolated reconstructed image sequence to generate a smooth visual motion image sequence. Visual coherence is maintained between switching frames. Insert calculated intermediate frames between each frame to improve the smoothness of the animation.

[0030] In this embodiment, step S4 includes the following steps: Calculate the mirror curvature information of the car rearview mirror; Constructing a surface projection inverse mapping model according to the mirror curvature information; Identifying the real-time eye position of the driver, performing line of sight simulation mapping on a curved surface projection inverse mapping model based on the real-time eye position, and identifying a reflection point of the mapping model; Traverse all the model reflection points and build the inverse transformation mapping table of the mirror reflection position; According to the inverse transformation mapping table, the mirror deformation correction and pixel relocation mapping are performed on the coherent visual animation sequence to construct the distortion-corrected animation sequence.

[0031] In this embodiment, a high-precision laser scanner is used to scan the surface of a car rearview mirror, acquiring detailed three-dimensional point cloud data. This data will be used for subsequent curvature calculations. During the laser scanning process, the scanning frequency is set to 1000 Hz to ensure that a large amount of point cloud data is acquired in a short period of time, covering the entire mirror surface. The acquired point cloud data is processed, and curvature information is calculated using a surface fitting algorithm (such as the least squares method). The curvature values at different locations are determined by analyzing the surface normal vector. The calculated curvature radius of the mirror surface is 500 mm at the center and 700 mm at the edge, forming a mirror curvature model. Based on the calculated mirror curvature information, a surface projection inverse mapping model is constructed. This model should be able to project the observer's line of sight onto the mirror surface and calculate the corresponding reflection point. Based on the principles of geometric optics, the model can calculate the reflection angle using the mirror normal and the angle of incidence. When constructing the model, the necessary geometric parameters are set, including the mirror normal vector, the position of the observation point, and the angle of incidence of the line of sight. This ensures that the model accurately reflects the optical properties of the mirror surface. Set the initial observation point distance from the mirror to 1 meter and the angle of incidence to 30 degrees. Calculate the corresponding reflection point location. Verify and adjust the inverse mapping model by comparing the actual reflected light with the model's calculated results. Ensure the model's accuracy for subsequent gaze simulation mapping. If the model's calculated reflection point deviates from the actual observation, adjust the model's parameters to ensure the calculated reflection point is consistent with the actual situation. Select appropriate eye tracking technology to identify the driver's eye position in real time. Common technologies include infrared eye trackers and video eye tracking. Infrared eye tracking devices can accurately capture the driver's eye movement trajectory and position. Collect the driver's eye position data in real time in the driving environment. Ensure that the device's sensitivity and sampling frequency meet the requirements of real-time monitoring. Set the device's sampling frequency to 60Hz to record eye position changes in real time and ensure data accuracy. Input the real-time eye position data into the constructed surface projection inverse mapping model for gaze simulation mapping. The model will calculate the corresponding reflection point based on the eye position. If the real-time eye position is (300, 400), the model calculates the reflection point to be (320, 420), ensuring that the calculation of the reflected light complies with the principles of geometric optics. Record the calculated reflection points and generate a list containing all reflection points. This will provide basic data for the subsequent construction of the inverse transformation mapping table. Record the reflection point array as [(320, 420), (310, 410), ...] for subsequent traversal and analysis. Define the method for constructing the inverse transformation mapping table. This mapping table should be able to correspond the reflection points with the pixel positions in the actual image to facilitate subsequent distortion correction. Use bilinear interpolation to map the coordinates of the reflection points to the pixel coordinates of the image. Traverse all calculated reflection points, calculate the corresponding pixel position of each reflection point in the image in turn, and record them in the inverse transformation mapping table.Correct mirror distortion for a coherent visual animation sequence based on the inverse transformation mapping table. Use remapping technology to reposition each pixel according to the mapping table. Use image resampling technology to ensure that each pixel is adjusted according to its inverse transformation position. Process each frame of the visual animation sequence and apply the inverse transformation mapping table to reposition the pixels. Ensure that each frame of the image is corrected to eliminate mirror distortion. For each frame of the image, calculate the new position of each pixel according to the mapping table and generate a corrected image frame. Record the corrected image sequence to generate a motion sequence containing all corrected frames for subsequent visualization and analysis. Record the generated corrected motion sequence for subsequent display in the car's rearview mirror.

[0032] In this embodiment, the specific steps of step S5 are: Calculate the multi-period position of the moving object in the distortion-corrected moving image sequence to obtain the multi-period movement trajectory; Calculating the moving speed and the speed change rate according to the multi-period moving trajectory; Performing direction change frequency distribution analysis on the multi-period movement trajectory to extract direction change frequency distribution information; Performing behavior pattern analysis based on the movement speed, speed change rate, and direction change frequency distribution information to generate a personalized behavior pattern for the moving object; Behavior trajectory prediction is performed based on the personalized behavior pattern of the moving object to generate the predicted behavior trajectory of the moving object.

[0033] In this embodiment, we define how to extract the multi-time interval positions of moving objects from a distortion-corrected animated image sequence. Object detection algorithms (such as YOLO or SSD) can be used to identify and track moving objects in the image. A target detection model is configured to identify moving objects such as vehicles and pedestrians, ensuring that the model works effectively under different lighting conditions and viewing angles. The corrected animated image sequence is processed frame by frame, and the target detection model is invoked to detect and record the positions of moving objects in each frame. The recorded positions should include the object's bounding box coordinates and its category. If a vehicle is detected in a frame with bounding box coordinates (x1, y1, x2, y2), this information is recorded as a combination of a timestamp and position, such as (t, (x1, y1, x2, y2)). A method for calculating movement speed is defined. Speed can be obtained by calculating the distance and time interval between adjacent position points. The recorded multi-time interval positions are processed frame by frame, and the movement speed within each time interval is calculated and recorded. Ensure that the speed units are consistent (e.g., meters per second). The rate of change of speed, i.e., the change in speed between adjacent time intervals, is calculated. A method for calculating direction change is defined. Direction can be obtained by calculating the angular change between adjacent locations. A common calculation method is to use the inverse tangent function. The recorded movement positions over multiple time periods are processed frame by frame. The directional change between each pair of adjacent locations is calculated and the angle of change is recorded. The calculated directional changes are statistically analyzed to generate a frequency distribution of directional changes. This can be achieved using a histogram to analyze the frequency of different angular changes. The frequency of angular changes within intervals such as 0-10 degrees and 10-20 degrees is counted to plot a directional change frequency distribution. Define the behavior patterns of moving objects and construct a behavior pattern recognition model based on speed, speed change rate, and direction change frequency. Clustering algorithms or machine learning models (such as support vector machines) can be used for behavior classification. Set the clustering algorithm to K-means and cluster the extracted features to identify different behavior patterns (such as fast driving and sharp turns). Extract features from the extracted speed, speed change rate, and direction change frequency as input data for behavior pattern analysis. Ensure that the features reflect the object's motion state. Extract the feature vector [v, Δv, direction change frequency] and input it into the model for analysis. Based on the identified personalized behavior patterns, a behavior trajectory prediction model is constructed. Future trajectories can be predicted using a time series prediction model (such as LSTM) or linear regression. The LSTM model input is set as feature data from the previous n frames to predict the trajectory for the next n frames. Historical behavior patterns and feature values are input into the constructed prediction model to generate a predicted trajectory for the moving object. The prediction results should be multi-time period location points. The generated predicted trajectory of the moving object is recorded to generate time series data containing the predicted location points for subsequent analysis and visualization.

[0034] In this embodiment, the specific steps of step S6 are: Obtain the real-time operating conditions of the vehicle; infer the driver's driving intention based on the real-time operating conditions of the vehicle and extract the driving intention inference data; Calculating the collision risk of the predicted trajectory of the moving object based on the driving intention inference data to obtain a collision risk level of the moving object; Based on a preset risk threshold, the collision risk level of the moving object is marked as a high-risk target to obtain a high-risk target object; The high-risk target object is subjected to highlight visualization processing on the distortion-corrected dynamic image sequence to perform a rearview mirror dynamic image projection operation.

[0035] In this embodiment, a method for acquiring real-time operating conditions is defined. Vehicle operating conditions can be acquired through the OBD-II (Onboard Diagnostics) interface, including parameters such as speed, rotational speed, accelerator pedal position, and brake status. This data is acquired in real time by connecting to the vehicle's ECU (Electronic Control Unit) via an OBD-II adapter. The data acquisition frequency is set to 1 Hz to ensure data timeliness and accuracy. While the vehicle is in motion, a data acquisition device (such as an OBD-II adapter) records the vehicle's operating condition data in real time. This data is used for subsequent driving intention inference. Recorded data includes current speed of 50 km / h, engine speed of 3000 rpm, accelerator pedal position of 40%, and brake status as inactive. The collected operating condition data is organized into a data log table. This ensures easy access and analysis of the data for subsequent driving intention inference. Based on the real-time operating condition data, an appropriate driving intention inference model is selected. Machine learning-based models (such as decision trees and random forests) or rule-based inference systems can be used. The model is configured to identify the driver's intentions, such as acceleration, deceleration, and steering. The model input is collected driving condition data, and its output is driving intention. When training the model, historical datasets, including labeled driving intention data, are required so that the model can effectively learn driving behavior under different driving conditions. If the current throttle position is 40% and the vehicle speed is 50 km / h, the model may infer the driver's intention to "accelerate" or "maintain speed." The inferred driving intention is recorded to generate a table of inferred driving intention data. Ensure the data format is standardized for subsequent analysis. Based on the inferred driving intention data, a collision risk calculation method is defined. Collision risk can be calculated by evaluating the relative position and velocity between the predicted trajectory of a moving object and the driving intention. The risk assessment formula is set as R = d / t, where d is the minimum distance between the predicted trajectory and the vehicle, and t is the warning time. The predicted trajectory of each moving object is analyzed and combined with the inferred driving intention data to calculate the collision risk level for each object. Ensure that the calculation process accurately reflects potential collision risks. A preset collision risk threshold is set to identify high-risk targets. The threshold setting should be adjusted based on the actual application scenario and safety standards. Set the threshold to 2.0, indicating that objects with a risk level exceeding this value are considered high risk. Traverse the conflict risk levels of all moving objects and mark them according to the set threshold. Identify high-risk target objects for subsequent processing. If the risk level of object 1 is 2.5, which exceeds the threshold, it will be marked as a high-risk target, and the result is recorded as selecting an appropriate visualization processing method to highlight the high-risk target object. Image overlay technology can be used to highlight high-risk targets in the animation sequence. Set the display color of high-risk targets to red so that they are easy to identify in the image. Process the distortion-corrected animation sequence of each frame, traverse the high-risk target objects, and apply a highlighting effect.Add highlight marks to the borders of each target object. Record and display the processed animated image sequence to ensure that high-risk targets are clearly visualized to improve driving safety.

[0036] The generated animated sequence will be played in the rearview mirror at 30 frames per second, and the highlighted objects will be highlighted with a red border, ensuring that the driver can promptly notice high-risk situations. [(Object 1, High Risk)].

[0037] In this embodiment, a motion picture projection system based on a car rearview mirror is provided, which is used to execute the above-mentioned motion picture projection method based on a car rearview mirror, including: Light interference optimization module, used to obtain the car rear view panoramic multispectral image stream, and perform adaptive ambient light interference optimization to generate a multi-level ambient light interference optimized image stream; An image processing module is used to perform dynamic light intensity change adaptation calculation and nonlinear gamma correction on the multi-level ambient light interference optimized image stream to generate a dynamic light adaptation image frame sequence; The inter-frame pixel reconstruction module is used to perform reverse dynamic disparity prediction on the dynamic light adaptation image frame sequence and perform inter-frame pixel migration reconstruction to construct a coherent visual animation sequence; The distortion correction module is used to calculate the mirror curvature information of the car's rearview mirror, and perform mirror deformation correction and pixel relocation mapping on a coherent visual animation sequence to construct a distortion-corrected animation sequence; The trajectory prediction module is used to calculate the multi-time position of the moving object and predict the behavior trajectory based on the distortion-corrected dynamic image sequence to generate the predicted behavior trajectory of the moving object; The highlight visualization module is used to calculate collision risk based on the predicted trajectory of moving objects, mark high-risk target objects, and highlight and visualize the distortion-corrected animation sequence.

[0038] This invention utilizes an ambient light interference optimization mechanism to identify strong light, reflections, and high-brightness background interference (such as high beams and direct sunlight) in real time, dynamically masking and balancing it. Its multi-layered image output provides a multi-dimensional reference image source for subsequent image processing, improving the system's robustness to interference and ensuring image stability and continuity. A dynamic adaptation algorithm aligns local brightness and overall illumination across consecutive image frames, ensuring brightness consistency between images and avoiding visual jumps. Gamma correction enhances dark detail and highlight suppression, improving image contrast and texture clarity. This provides a unified, stable, and faithfully reproduced image sequence, providing an ideal input source for animated image construction and enhancing rearview mirror visual clarity and responsiveness. A disparity prediction algorithm identifies relative depth changes between moving objects and backgrounds within image frames, enabling accurate image temporal vector modeling. Pixel migration technology generates fitted frames from non-consecutive frames, effectively eliminating animation artifacts such as stuttering, ghosting, and frame skipping. The resulting visual animation is coherent and natural, creating an augmented reality-like rearview animation experience and improving the driver's spatial judgment and attention acquisition. The system accurately captures the mirror's actual surface parameters and combines them with a dynamic image projection deformation model to achieve precise image geometry correction. A pixel repositioning algorithm ensures that the projected image closely corresponds to the actual physical location, preventing object position misjudgment due to distortion. The corrected dynamic image projection eliminates parallax errors without changing traditional rearview mirror usage, achieving a hybrid visual experience combining "real location + information enhancement." Multi-period behavioral modeling is performed based on the target's position sequence, speed variation, and trajectory curvature within the dynamic image, enabling high-precision prediction of dynamic objects. The system accommodates a wide range of targets, including vehicles, bicycles, pedestrians, small animals, and other dynamic entities of varying sizes. Predicted trajectories serve as a prerequisite for behavioral risk assessment, laying the foundation for proactive system responses (such as target highlighting and voice warnings). Collision risk calculation analyzes potential collision trends based on relative motion vectors, angular proximity, and speed differences. High-risk targets are visually highlighted in the dynamic image through color enhancement, outlines, and dynamic icons, creating a head-up display (HUD)-like visual prompt. This effectively directs the driver's attention to potential hazards, improving driving safety, particularly in complex urban traffic conditions or high-speed lane changes.

[0039] 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.

[0040] 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 method for projecting a moving picture based on a car rearview mirror, characterized in that: The following steps are involved: Step S1: Obtaining a vehicle rearview panoramic multispectral image stream and performing adaptive ambient light interference optimization to generate a multi-level ambient light interference optimized image stream; Step S2: performing dynamic light intensity change adaptation calculation and nonlinear gamma correction on the multi-level ambient light interference optimized image stream to generate a dynamic light adaptation image frame sequence; Step S3: performing reverse dynamic disparity prediction on the dynamic light adaptation image frame sequence and performing inter-frame pixel migration reconstruction to construct a coherent visual animation sequence; Step S4: Calculate the mirror curvature information of the rearview mirror of the car, and perform mirror deformation correction and pixel relocation mapping on the coherent visual animation sequence to construct a distortion-corrected animation sequence; Step S5: Calculate the position of the moving object in multiple time periods and predict its behavior trajectory based on the distortion-corrected moving image sequence to generate a predicted behavior trajectory of the moving object; Step S6: Calculate the collision risk based on the predicted trajectory of the moving object, mark high-risk target objects, and highlight and visualize the distortion-corrected animated image sequence.

2. The method for projecting a moving picture based on a car rearview mirror according to claim 1, characterized in that: The specific steps of step S1 are: Based on the low-light multispectral camera unit, high dynamic range real-time sampling is performed to obtain the vehicle rear view panoramic multispectral image stream; Performing multi-level light characteristic analysis on the vehicle rearview panoramic multispectral image stream to identify the ambient light color temperature, light intensity, and dispersion; Performing ambient light scene analysis based on the ambient light color temperature, light intensity, and dispersion to identify the current ambient light scene; Adaptively optimize the ambient light interference of the vehicle rear-view panoramic multi-spectral image stream according to the ambient light scene to generate a multi-level ambient light interference optimized image stream.

3. The method for projecting a moving image based on a car rearview mirror according to claim 2, characterized in that: The step of adaptively optimizing the ambient light interference of the vehicle rearview panoramic multispectral image stream according to the ambient light scene to generate a multi-level ambient light interference optimized image stream is as follows: Performing illumination interference blur calculation on the ambient light scene to identify the interference blur degree of the current scene; performing adaptive light compensation analysis according to the interference blur degree to obtain light interference compensation parameters; Identify the ambient light interference area of the car rearview panoramic multispectral image stream and extract the sub-image of the light interference area; Adaptively suppress and optimize the sub-image of the light interference area according to the light interference compensation parameter to generate a light interference suppression optimized sub-image; The vehicle rearview panoramic multispectral image stream is globally fused and reconstructed based on the light interference suppression optimized sub-image to generate a multi-level ambient light interference optimized image stream.

4. The method for projecting a moving picture based on a car rearview mirror according to claim 1, wherein: The specific steps of step S2 are: Perform global frame-by-frame decomposition of the multi-level ambient light interference optimized image stream to generate a time-series frame image sequence; Calculating the average brightness value of each frame image in the time-sequential frame image sequence; Calculating the brightness difference between frames based on the average brightness value to generate brightness difference information between frames; Performing an ambient light intensity fluctuation trend analysis on the current ambient light blur scene to extract ambient light intensity fluctuation trend characteristics; Based on the characteristics of the ambient light intensity fluctuation trend, the inter-frame brightness difference signal is dynamically adapted to calculate the brightness change to obtain the inter-frame brightness video compensation value; The multi-level ambient light interference optimized image stream is subjected to nonlinear gamma correction according to the inter-frame brightness video compensation value to generate a dynamic light adaptation image frame sequence.

5. The method for projecting a moving image based on a car rearview mirror according to claim 1, wherein: The specific steps of step S3 are: Perform deep visual recognition on the dynamic light adaptation image frame sequence and perform convolutional network semantic segmentation to extract static obstacles and moving objects in the scene; Perform dynamic optical flow tracking and frame-by-frame position calculation on moving objects to obtain multi-frame pixel position points of moving objects; Performing multi-time point disparity calculation based on the multi-frame pixel position points to generate multi-frame object disparity; Calculating a temporal parallax change rate of the multi-frame object parallax; Perform reverse dynamic disparity prediction based on the temporal disparity change rate to generate a reverse dynamic disparity difference value; Based on the inverse dynamic parallax difference, the dynamic light adaptation image frame sequence is reconstructed by inter-frame pixel migration to generate a dynamic interpolation reconstructed image; Perform temporal motion graphics fitting based on the dynamic interpolation reconstructed images to construct a coherent visual motion graphics sequence.

6. The method for projecting a moving picture based on a car rearview mirror according to claim 1, wherein: The specific steps of step S4 are: Calculate the mirror curvature information of the car rearview mirror; Constructing a surface projection inverse mapping model according to the mirror curvature information; Identifying the real-time eye position of the driver, performing line of sight simulation mapping on a curved surface projection inverse mapping model based on the real-time eye position, and identifying a reflection point of the mapping model; Traverse all the model reflection points and build the inverse transformation mapping table of the mirror reflection position; According to the inverse transformation mapping table, the mirror deformation correction and pixel relocation mapping are performed on the coherent visual animation sequence to construct the distortion-corrected animation sequence.

7. The method for projecting a moving image based on a car rearview mirror according to claim 1, wherein: The specific steps of step S5 are: Calculate the multi-period position of the moving object in the distortion-corrected moving image sequence to obtain the multi-period movement trajectory; Calculating the moving speed and the speed change rate according to the multi-period moving trajectory; Performing direction change frequency distribution analysis on the multi-period movement trajectory to extract direction change frequency distribution information; Performing behavior pattern analysis based on the movement speed, speed change rate, and direction change frequency distribution information to generate a personalized behavior pattern for the moving object; Behavior trajectory prediction is performed based on the personalized behavior pattern of the moving object to generate the predicted behavior trajectory of the moving object.

8. The method for projecting a moving image based on a car rearview mirror according to claim 1, characterized in that: The specific steps of step S6 are: Obtain the real-time operating conditions of the vehicle; infer the driver's driving intention based on the real-time operating conditions of the vehicle and extract the driving intention inference data; Calculating the collision risk of the predicted trajectory of the moving object based on the driving intention inference data to obtain a collision risk level of the moving object; Based on a preset risk threshold, the collision risk level of the moving object is marked as a high-risk target to obtain a high-risk target object; The high-risk target object is subjected to highlight visualization processing on the distortion-corrected dynamic image sequence to perform a rearview mirror dynamic image projection operation.

9. A motion picture projection system based on a car rearview mirror, characterized in that: The method for executing the dynamic image projection method based on the automobile rearview mirror according to claim 1 comprises: Light interference optimization module, used to obtain the car rear view panoramic multispectral image stream, and perform adaptive ambient light interference optimization to generate a multi-level ambient light interference optimized image stream; An image processing module is used to perform dynamic light intensity change adaptation calculation and nonlinear gamma correction on the multi-level ambient light interference optimized image stream to generate a dynamic light adaptation image frame sequence; The inter-frame pixel reconstruction module is used to perform reverse dynamic disparity prediction on the dynamic light adaptation image frame sequence and perform inter-frame pixel migration reconstruction to construct a coherent visual animation sequence; The distortion correction module is used to calculate the mirror curvature information of the car's rearview mirror, and perform mirror deformation correction and pixel relocation mapping on a coherent visual animation sequence to construct a distortion-corrected animation sequence; The trajectory prediction module is used to calculate the multi-time position of the moving object and predict the behavior trajectory based on the distortion-corrected dynamic image sequence to generate the predicted behavior trajectory of the moving object; The highlight visualization module is used to calculate collision risk based on the predicted trajectory of moving objects, mark high-risk target objects, and highlight and visualize the distortion-corrected animation sequence.

Images