Omnibearing blind area monitoring and driving safety auxiliary positioning system

Abstract

The invention discloses an omnibearing blind area monitoring and driving safety auxiliary positioning system, which relates to the field of intelligent traffic systems and specifically comprises a blind area monitoring module, a depth point cloud module, a target tracking module, a target prediction module and an AR display module. High-resolution cameras are installed at all parts of a vehicle, environment image data are collected in real time, feature point detection and matching are carried out, a depth map and left and right environment point cloud data are generated, optical flow analysis and the depth map are utilized, target changes are dynamically detected, real-time monitoring is ensured, a deep learning model is constructed, and the real-time monitoring is realized. Advanced features are extracted, time sequence analysis is carried out, the target motion position is predicted, whether the target enters a monitoring blind area is judged, alarm information is visualized through the augmented reality technology, multiple warning lamps and color codes are used, target motion and the vehicle position and direction are visually indicated, and therefore the driving safety is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation systems, and particularly to an all-round blind area monitoring and driving safety assistance positioning system. Background Art

[0002] With the rapid progress in the fields of intelligent transportation and automated driving, the requirements for vehicle safety and driving assistance systems have become increasingly important. The complexity and variability of the modern traffic environment make it necessary for drivers to rely on more advanced technologies to improve driving safety and reduce the probability of accidents. Against this background, the demand for comprehensively monitoring and evaluating the vehicle's surrounding environment has been increasing day by day, which has promoted the research and application of all-round blind area monitoring and driving safety assistance positioning systems.

[0003] The existing technologies have the following deficiencies: The existing vehicle blind area monitoring systems mainly rely on single or limited sensor configurations and only cover certain specific areas of the vehicle, often unable to comprehensively identify the blind areas around the vehicle, resulting in drivers being unable to detect potential dangers in a timely manner when changing lanes, reversing, or performing other operations; in addition, the existing target detection technologies mostly rely on simple image processing algorithms, making it difficult to adapt to complex and changeable driving environments, resulting in insufficient accuracy of target detection, lacking the prediction of target movement, and being unable to identify potential dangers in advance, posing safety hazards to drivers. The single way of transmitting alarm information is difficult to quickly attract the driver's attention and lacks user-defined functions, unable to meet the personalized needs of different drivers.

[0004] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and thus it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide an all-round blind area monitoring and driving safety assistance positioning system to solve the problems in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solutions: An all-round blind area monitoring and driving safety assistance positioning system, specifically including: a blind area monitoring module, a depth point cloud module, a target tracking module, a target prediction module, and an AR display module; Blind area monitoring module: Divide different parts of the vehicle into different monitoring blind areas and respectively install two high-resolution cameras to collect left and right environmental image data in real time; Depth point cloud module: Generate a depth map and left and right environmental point cloud data by performing feature point detection, matching, and calculating the disparity value on the left and right environmental image data; Target tracking module: Dynamically detect target changes by analyzing the optical flow of consecutive frames of the left and right environmental image data and combining with the depth map using a target tracker; Target prediction module: Build a deep learning model, extract high-level features based on environmental point cloud data and target motion information, perform time series analysis, predict the motion position of the target, and determine whether it enters the monitoring blind area of the vehicle; AR display module: Visualize the alarm information through augmented reality technology, design various forms of warning lights and combine color coding to achieve warnings for target motion, indication of vehicle position and direction; Preferably, divide and label the front part of the vehicle as the first monitoring blind area, the side part of the vehicle as the second monitoring blind area, and the rear part of the vehicle as the third monitoring blind area. Install two high-resolution cameras at the front, side, and rear of the vehicle respectively to collect left and right environmental image data in real time. The left and right environmental image data includes static target data and dynamic target data. Mark the timestamp of each environmental image data, keep the monitoring angle of the high-resolution camera at 45° with the target angle, and calibrate the spatial position of the high-resolution camera in the vehicle environment and the overlap of the monitoring angles.

[0007] Preferably, perform grayscale processing on the left and right environmental image data, use feature point detection and description algorithms to find the matching feature points of the target in the left and right environmental image data, determine the corresponding relationship of the matching feature points of each target in the left and right environmental image data through the KNN matching algorithm, calculate the horizontal disparity value of each pixel in the left and right environmental image data and create a disparity map. The specific formula is:

[0008] where, represents the horizontal disparity value of each pixel in the left and right environmental image data, represents the horizontal coordinate of the matching feature point of the target in the left environmental image data, represents the horizontal coordinate of the matching feature point of the target in the right environmental image data. Calculate the depth value of each pixel in the left and right environmental image data based on the horizontal disparity value and create a depth map. The specific formula is:

[0009] where, represents the depth value of each pixel in the left and right environmental image data, represents the focal length of the high-resolution camera, represents the baseline of the high-resolution camera. Calculate the three-dimensional coordinates of each pixel in the left and right environmental image data based on the depth map and generate left and right environmental point cloud data. The specific formula is:

[0010] where, represents the abscissa of the left and right environmental point cloud data, represents the vertical coordinates of the left and right environmental point cloud data, represents the two-dimensional coordinates of each pixel in the left and right environmental image data, represents the principal point coordinates of the high-resolution camera.

[0011] Preferably, the motion of the target is estimated by analyzing the pixel intensity change based on consecutive frames of the left and right environmental image data. The optical flow is calculated for consecutive frames, and its specific formula is:

[0012] where, represents the left and right environmental point cloud data optical flow component in the direction, represents the left and right environmental point cloud data optical flow component in the direction, represents the gradient in the time direction of the left and right environmental point cloud data, represents the regularization parameter that controls the smoothness of the optical flow field. The stationary state before the vehicle starts is selected as the initial vehicle background. The difference between the current frame of the environmental image data and the corresponding frame of the initial vehicle background is compared to achieve dynamic and adaptive target change detection. Targets that are too far or too close to the high-resolution camera are filtered according to the depth value of each pixel in the left and right environmental image data. A bounding box is created to detect the vehicle background and initialize the target tracker. A set of particles representing the possible states of the target is randomly generated near the initial position of the target and the same initial weight is assigned. The positions of the particles are updated according to the acceleration motion principle, and the weights of each particle are updated based on the bounding box and the depth map. All weights are normalized and resampled, low-weight particles are removed, and high-weight particles are replicated.

[0013] Preferably, a deep learning model is constructed to receive left and right environmental point cloud data and target motion information through an input layer, add multiple convolutional layers, use ReLU as an activation function layer to introduce non-linearity and summarize and extract environmental point cloud features and target motion features through convolutional kernels, perform downsampling operations on environmental point cloud features and target motion features through a pooling layer, repeat the first convolution and pooling operations iteratively to extract high-level features of left and right environmental point cloud data, add a batch normalization layer to each convolutional layer to normalize environmental point cloud features and target motion features in the convolutional kernels, combine environmental point cloud features and target motion features with timestamps to form a time series through an LSTM layer, receive the time series through an input gate, and gradually pass the time series to the LSTM layer. At each time step, the LSTM layer receives the input environmental point cloud features, target motion features of the current time step and the hidden state of the previous time step, and generates the output environmental point cloud features, target motion features of the current time step and the hidden state of the current time step. An output time series feature is generated by each LSTM layer in the long short-term memory network, where the output of each time step contains the time series information of that time step. The output time series feature of the LSTM is received through a fully connected layer and the predicted motion position of the target is mapped and output. Whether the predicted motion position of the target enters the monitoring blind area of the vehicle background is judged according to the difference value of each pixel in the left and right environmental image data, and the position of the high-resolution camera is associated to judge the part of the vehicle to which the current monitoring blind area belongs.

[0014] Preferably, the alarm information is visualized on the in-vehicle display through augmented reality technology. A triangular warning light is designed to represent target motion warnings, a circular warning light is designed to represent vehicle position warnings, and an arrow warning light is designed to represent target direction warnings. The warning lights are color-coded, where a red warning light represents a high risk, a yellow warning light represents a warning, a green warning light represents a safe state, and a blue warning light represents additional road information. Periodic flashing of the warning lights and smooth movement of the target are added. The coordinates of the left and right environmental point cloud data are converted into the coordinates of the in-vehicle display. According to the monitoring perspectives and depth maps of different high-resolution cameras, the size and position of the augmented reality content are adjusted using perspective projection, allowing the driver to customize the alarm level and the display content method.

[0015] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. By precisely dividing the monitoring blind spots of the vehicle and installing high-resolution cameras in each blind spot, it is possible to collect left and right environmental image data in real time, significantly improving the perception ability of the surrounding environment. Through grayscale processing and feature point detection of the environmental image data, the system can effectively identify static and dynamic targets, and establish matching feature points of the targets through the KNN matching algorithm, which not only improves the accuracy of target detection but also lays a foundation for subsequent depth value calculation and three-dimensional coordinate generation. By using optical flow analysis technology to perform motion estimation on consecutive frames, the system can dynamically monitor the motion changes of targets, promptly identify potential dangerous situations, and make target tracking more accurate through the particle filter method. By updating and resampling the particle weights, the system can adapt to complex environmental changes and ensure stable tracking of the target state, greatly enhancing the real-time response ability of the system in complex driving scenarios and providing more comprehensive safety protection for drivers.

[0016] 2. By inputting left and right environmental point cloud data and target motion information, using convolutional neural networks and long short-term memory networks to extract high-level features and form time series analysis, it not only improves the prediction ability of the target motion state but also can learn and adapt to different driving environments, can judge the predicted motion position of the target in real time, and determine whether it enters the monitoring blind spot of the vehicle. By visualizing the alarm information through augmented reality technology, the system can intuitively present target motion warnings, vehicle position warnings, and target direction warnings on the in-vehicle display screen, improving the driver's alertness and reaction speed. By designing color-coded warning lights, the information transmission is made clearer, and the driver's attention can be quickly guided. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0018] Figure 1 It is a flowchart of the method for the all-round blind spot monitoring and driving safety auxiliary positioning system of the present invention.

[0019] Figure 2 It is a schematic diagram of the modules of the all-round blind spot monitoring and driving safety auxiliary positioning system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art.

[0021] The present invention provides an Figure 2 omnidirectional blind spot monitoring and driving safety assisted positioning system as shown, specifically including: a blind spot monitoring module, a depth point cloud module, a target tracking module, a target prediction module, and an AR display module; Blind spot monitoring module: Different parts of the vehicle are divided into different monitoring blind spots, and two high-resolution cameras are respectively installed to collect left and right environmental image data in real time; The front part of the vehicle is divided and marked as the first monitoring blind spot, the side part of the vehicle is divided and marked as the second monitoring blind spot, and the rear part of the vehicle is divided and marked as the third monitoring blind spot. Two high-resolution cameras are respectively installed at the front, side, and rear of the vehicle to collect left and right environmental image data in real time. The left and right environmental image data includes static target data and dynamic target data. The timestamp of each environmental image data is marked, and the monitoring angle of the high-resolution camera is kept at 45° with the target angle, and the spatial position of the high-resolution camera in the vehicle environment and the overlap of the monitoring angles are calibrated.

[0022] Depth point cloud module: By performing feature point detection, matching, and calculating the disparity value on the left and right environmental image data, a depth map and left and right environmental point cloud data are generated; The left and right environmental image data is grayscale processed, and the feature point detection and description algorithm is used to find the matching feature points of the targets in the left and right environmental image data. Among them, the feature point detection and description algorithms include SIFT, SURF, and ORB. The corresponding relationship of the matching feature points of each target in the left and right environmental image data is determined through the KNN matching algorithm, and the horizontal disparity value of each pixel in the left and right environmental image data is calculated and a disparity map is created. The specific formula is:

[0023] where, represents the horizontal disparity value of each pixel in the left and right environmental image data, represents the horizontal coordinate of the matching feature point of the target in the left environmental image data, represents the horizontal coordinate of the matching feature point of the target in the right environmental image data. The depth value of each pixel in the left and right environmental image data is calculated according to the horizontal disparity value and a depth map is created. The specific formula is:

[0024] where, Represents the depth value of each pixel in the left and right environment image data, represents the focal length of the high-resolution camera, Represents the baseline of the high-resolution camera. Based on the depth map, the three-dimensional coordinates of each pixel in the left and right environmental image data are calculated and the left and right environmental point cloud data are generated. The specific formula is:

[0025] in, Represents the horizontal coordinates of the left and right environmental point cloud data, Indicates the vertical coordinates of the left and right environmental point cloud data. Represents the two-dimensional coordinates of each pixel in the left and right environmental image data, Represents the principal point coordinates of the high-resolution camera.

[0026] Target tracking module: By analyzing the optical flow of consecutive frames of left and right environmental image data, the target tracker is used to dynamically detect target changes in combination with the depth map; The target motion is estimated by analyzing the pixel intensity changes in the continuous frames of the left and right environmental image data, and the optical flow is calculated for the continuous frames. The specific formula is:

[0027] in, Represents the left and right environmental point cloud data Directional optical flow component, Represents the left and right environmental point cloud data Directional optical flow component, Represents the gradient of the left and right environmental point cloud data in the time direction, It represents the regularization parameter that controls the smoothness of the optical flow field. The static state of the vehicle before starting is selected as the initial vehicle background. The difference between the current frame of the environmental image data and the corresponding frame of the initial vehicle background is compared to realize dynamic adaptive target change detection. According to the depth value of each pixel in the left and right environmental image data, the targets that are too far or too close to the high-resolution camera are filtered out. A bounding box is created to detect the vehicle background and initialize the target tracker. A group of particles representing the possible states of the target are randomly generated near the initial position of the target and assigned the same initial weight. The position of the particles is updated according to the principle of acceleration motion, and the weight of each particle is updated based on the bounding box and depth map. All weights are normalized and resampled, low-weight particles are eliminated, and high-weight particles are copied.

[0028] Target prediction module: Build a deep learning model, extract advanced features based on environmental point cloud data and target motion information, and perform time series analysis to predict the target's motion position and determine whether it enters the vehicle's monitoring blind spot; Build a deep learning model that receives left and right environmental point cloud data and target motion information through the input layer, adds multiple convolutional layers, uses ReLU as the activation function layer to introduce non-linearity and summarizes and extracts environmental point cloud features and target motion features through convolutional kernels, performs downsampling operations on environmental point cloud features and target motion features through pooling layers, repeats the first convolution and pooling operations iteratively to extract high-level features of left and right environmental point cloud data, adds batch normalization layers to each convolutional layer to normalize environmental point cloud features and target motion features in the convolutional kernels, combines environmental point cloud features and target motion features with timestamps to form a time series through the LSTM layer, where each time step represents environmental point cloud features and target motion features extracted by a recurrent neural network, receives the time series through the input gate, and gradually passes the time series to the LSTM layer. In each time step, the LSTM layer receives the input environmental point cloud features and target motion features of the current time step and the hidden state of the previous time step, and generates the output environmental point cloud features and target motion features of the current time step and the hidden state of the current time step. A time series feature of the output is generated through each LSTM layer in the long short-term memory network, where the output of each time step contains the time series information of that time step. The output time series feature of the LSTM is received through the fully connected layer and the predicted motion position of the target is mapped and output. Determine whether the predicted motion position of the target enters the monitoring blind area of the vehicle background based on the difference value of each pixel in the left and right environmental image data, and associate the position of the high-resolution camera to determine the part of the vehicle to which the current monitoring blind area belongs.

[0029] AR display module: Visualize the alarm information through augmented reality technology, design various forms of warning lights and combine color coding to achieve warnings for target motion, indication of vehicle position and direction; Visualize the alarm information on the in-vehicle display through augmented reality technology. Design a triangular warning light to indicate target motion warning, a circular warning light to indicate vehicle position warning, and an arrow warning light to indicate target direction warning. Color code the warning lights, where a red warning light indicates high risk, a yellow warning light indicates warning, a green warning light indicates a safe state, and a blue warning light indicates additional road information. Add periodic flashing of the warning lights and smooth movement of the target. That is, when the target moves towards the vehicle, the target image is smoothly moved on the in-vehicle display. Convert the coordinates of the left and right environmental point cloud data into the coordinates of the in-vehicle display. Adjust the size and position of the augmented reality content using perspective projection according to the monitoring perspectives and depth maps of different high-resolution cameras, and allow the driver to customize the alarm level and display content method.

[0030] By precisely dividing the monitoring blind spots of the vehicle and installing high-resolution cameras in each blind spot, it is possible to collect left and right environmental image data in real time, significantly improving the perception ability of the surrounding environment. Through grayscale processing and feature point detection of the environmental image data, the system can effectively identify static and dynamic targets, and establish matching feature points of the targets through the KNN matching algorithm, not only improving the accuracy of target detection, but also laying a foundation for subsequent depth value calculation and three-dimensional coordinate generation. By using optical flow analysis technology to perform motion estimation on consecutive frames, the system can dynamically monitor the motion changes of targets, timely identify potential dangerous situations, and make target tracking more accurate through the particle filter method. By updating and resampling the particle weights, the system can adapt to complex environmental changes and ensure stable tracking of the target state, greatly enhancing the real-time response ability of the system in complex driving scenarios and providing more comprehensive safety protection for the driver.

[0031] By inputting left and right environmental point cloud data and target motion information, using convolutional neural networks and long short-term memory networks to extract high-level features and form time series analysis, it not only improves the prediction ability of the target motion state, but also can learn and adapt to different driving environments, can judge the predicted motion position of the target in real time, and determine whether it enters the monitoring blind spot of the vehicle. By visualizing the alarm information through augmented reality technology, the system can intuitively present target motion warnings, vehicle position warnings, and target direction warnings on the in-vehicle display screen, improving the driver's alertness and reaction speed. By designing color-coded warning lights, the information transmission is made clearer, and the driver's attention can be quickly guided.

[0032] Only some exemplary embodiments of the present invention have been described above by way of illustration. Without doubt, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. All-round blind spot monitoring and driving safety auxiliary positioning system, characterized by: Specifically include: Including blind spot monitoring module, depth point cloud module, target tracking module, target prediction module and AR display module; Blind spot monitoring module: Different parts of the vehicle are divided into different monitoring blind spots, and two high-resolution cameras are installed to collect left and right environmental image data in real time; Depth point cloud module: generates depth map and left and right environment point cloud data by detecting and matching feature points and calculating disparity values ​​of left and right environment image data; Target tracking module: By analyzing the optical flow of consecutive frames of left and right environmental image data, the target tracker is used to dynamically detect target changes in combination with the depth map; Target prediction module: Build a deep learning model, extract advanced features based on environmental point cloud data and target motion information, and perform time series analysis to predict the target's motion position and determine whether it enters the vehicle's monitoring blind spot; AR display module: Visualize alarm information through augmented reality technology, design various forms of warning lights and combine them with color coding to achieve warnings on target movement and indication of vehicle position and direction.

2. The all-round blind spot monitoring and driving safety auxiliary positioning system according to claim 1 is characterized in that: In the blind spot monitoring module, different parts of the vehicle are divided into different monitoring blind spots, specifically including: the front of the vehicle is divided and marked as a first monitoring blind spot, the side of the vehicle is divided and marked as a second monitoring blind spot, and the rear of the vehicle is divided and marked as a third monitoring blind spot; two high-resolution cameras are respectively installed at the front, side and rear of the vehicle to collect left and right environmental image data in real time, wherein the left and right environmental image data include static target data and dynamic target data.

3. The all-round blind spot monitoring and driving safety auxiliary positioning system according to claim 1 is characterized in that: In the depth point cloud module, feature point detection, matching and calculation of disparity values ​​specifically include: using feature point detection and description algorithms to find matching feature points of targets in the left and right environmental image data, determining the corresponding relationship between the matching feature points of each target in the left and right environmental image data through the KNN matching algorithm, calculating the horizontal disparity value of each pixel in the left and right environmental image data and creating a disparity map.

4. The all-round blind spot monitoring and driving safety auxiliary positioning system according to claim 3 is characterized in that: The specific formula for calculating the horizontal disparity value of each pixel in the left and right environment image data is: ； in, Represents the horizontal disparity value of each pixel in the left and right environment image data. Represents the horizontal coordinates of the matching feature points of the target in the left environment image data, Represents the horizontal coordinates of the matching feature points of the target in the right environment image data.

5. The all-round blind spot monitoring and driving safety auxiliary positioning system according to claim 1 is characterized in that: In the depth point cloud module, generating the depth map and the left and right environment point cloud data specifically includes: calculating the depth value of each pixel in the left and right environment image data according to the horizontal disparity value and creating the depth map, and the specific formula is: ； in, Represents the depth value of each pixel in the left and right environment image data, represents the focal length of the high-resolution camera, Represents the baseline of the high-resolution camera. Based on the depth map, the three-dimensional coordinates of each pixel in the left and right environmental image data are calculated and the left and right environmental point cloud data are generated. The specific formula is: ； in, Represents the horizontal coordinates of the left and right environmental point cloud data, Indicates the vertical coordinates of the left and right environmental point cloud data. Represents the two-dimensional coordinates of each pixel in the left and right environmental image data, Represents the principal point coordinates of the high-resolution camera.

6. The all-round blind spot monitoring and driving safety auxiliary positioning system according to claim 1, characterized in that: In the target tracking module, the specific formula for the optical flow of consecutive frames of left and right environment image data is: ； in, Represents the left and right environmental point cloud data Directional optical flow component, Represents the left and right environmental point cloud data Directional optical flow component, Represents the gradient of the left and right environmental point cloud data in the time direction, Represents the regularization parameter that controls the smoothness of the optical flow field.

7. The all-round blind spot monitoring and driving safety auxiliary positioning system according to claim 1, characterized in that: In the target tracking module, the dynamic detection of target changes specifically includes: selecting the static state of the vehicle before starting as the initial vehicle background, comparing the difference between the current frame of the environmental image data and the corresponding frame of the initial vehicle background to realize dynamic adaptive target change detection, filtering the targets that are too far or too close to the high-resolution camera according to the depth value of each pixel in the left and right environmental image data, creating a bounding box to detect the vehicle background to initialize the target tracker, randomly generating a group of particles representing the possible states of the target near the initial position of the target, and assigning the same initial weight, updating the position of the particles according to the principle of acceleration motion, and updating the weight of each particle based on the bounding box and the depth map, normalizing all weights and resampling, eliminating low-weight particles, and copying high-weight particles.

8. The all-round blind spot monitoring and driving safety auxiliary positioning system according to claim 1, characterized in that: In the target prediction module, predicting the moving position of the target specifically includes: building a deep learning model, receiving the left and right environmental point cloud data and the target motion information through the input layer, adding multiple convolution layers, using ReLU as the activation function layer to introduce nonlinearity and summarizing and extracting the environmental point cloud features and the target motion features through the convolution kernel, downsampling the environmental point cloud features and the target motion features through the pooling layer, repeatedly iterating the first convolution and pooling operations to extract the high-level features of the left and right environmental point cloud data, adding a batch normalization layer to each convolution layer, normalizing the environmental point cloud features and the target motion features in the convolution kernel, and summarizing and extracting the environmental point cloud features and the target motion features through the convolution kernel. The features are combined with the timestamp to form a time series through the LSTM layer, the time series is received through the input gate, and the time series is gradually passed to the LSTM layer. In each time step, the input environment point cloud features, target motion features and the hidden state of the previous time step of the current time step are received through the LSTM layer, and the output environment point cloud features, target motion features and the hidden state of the current time step are generated. An output time series feature is generated through each LSTM layer in the long short-term memory network, where the output of each time step contains the time series information of the time step. The output time series features of the LSTM are received through the fully connected layer and the motion prediction position of the output target is mapped.

9. The all-round blind spot monitoring and driving safety auxiliary positioning system according to claim 1, characterized in that: In the AR display module, various forms of warning lights are combined with color coding, specifically including: designing a triangular warning light to indicate target motion warning, a circular warning light to indicate vehicle position warning, and an arrow warning light to indicate target direction warning. The warning lights are color-coded, wherein a red warning light indicates high risk, a yellow warning light indicates a warning, a green warning light indicates a safe state, and a blue warning light indicates additional road information. The warning lights are added with periodic flashing and smooth target movement.

Images