Automatic driving image processing system, method, equipment and medium

Abstract

The invention provides an automatic driving image processing system, method and device and a medium, and the system comprises an image real-time collection module which is used for collecting real-time image data through a sensor; the target detection module is used for extracting key features in the image through a feature extraction model, obtaining target features, detecting the target features and determining the position and the category of each target; the target recognition and tracking module is used for further recognizing the target, determining the specific type of the target and tracking the target through a target tracking algorithm; the scene understanding module is used for analyzing and determining traffic signals, signboards and lane line information according to detection and tracking results, understanding traffic rules and road conditions, generating text information and feeding back the text information to the decision-making module; the decision module is used for generating a vehicle driving decision according to the text information and the vehicle state; and the path planning module is used for planning the driving path of the vehicle at the future moment according to the driving decision and controlling the movement of the vehicle. According to the system, safe and efficient driving is ensured.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an automatic driving image processing system, method, device and medium. Background Art

[0002] Environmental perception is one of the most important functions of autonomous driving. The performance of environmental perception, such as accuracy, robustness to light changes and shadow noise, and adaptability to complex road environments and bad weather, directly affects the performance of autonomous driving technology. Sensors in autonomous driving include ultrasonic radar, millimeter wave radar, lidar, and visual sensors. Although global positioning technologies (such as GPS, Beidou, and GLONASS) are relatively mature and capable of all-weather positioning, there are still problems such as signal blocking or even loss, low update frequency, and poor positioning accuracy in environments such as urban buildings and tunnels.

[0003] The application of visual sensors in autonomous driving helps with object detection and image processing to analyze obstacles and drivable areas, thereby ensuring that the vehicle reaches its destination safely. Compared with other sensors, visual images, especially color images, contain a large amount of information. They not only contain distance information of objects, but also color, texture and depth information, so that lane line detection, vehicle detection, pedestrian detection, and traffic sign detection can be performed simultaneously through signal detection. Visual sensors also achieve simultaneous positioning and map construction.

[0004] However, traditional image processing methods of visual sensors, such as edge detection and Hough transform, have low algorithm complexity, but slow processing speed, limited accuracy and robustness. Traditional methods usually require customized hardware and software, complex system integration, and difficult maintenance. In view of this, there is an urgent need to provide an image processing system and method for autonomous driving to solve the above-mentioned problems. Summary of the invention

[0005] In order to overcome the problems existing in the related art, the present disclosure provides an autonomous driving image processing system, method, device and medium to solve the technical problems in the related art.

[0006] One or more embodiments of this specification provide an autonomous driving image processing system, including:

[0007] The real-time image acquisition module is used to acquire real-time image data through sensors;

[0008] Target detection module: used to extract key features in the image through the feature extraction model, obtain target features, and detect the extracted target features to determine the location and category of each target in the image;

[0009] Target recognition and tracking module: used to further identify the detected target, determine the specific type of the target, and track the target through the target tracking algorithm;

[0010] Scene understanding module: used to analyze and determine traffic signals, signs, lane line information, understand traffic rules and road conditions, and generate text information to feed back to the decision module based on the detection and tracking results of the target detection module and the target recognition and tracking module;

[0011] Decision module: used to generate vehicle driving decisions based on the text information and vehicle status of the scene understanding module; and

[0012] Path planning module: used to plan the vehicle's future driving path based on the driving decisions of the decision-making module and to control the movement of the vehicle.

[0013] Furthermore, an image preprocessing module is provided to perform denoising, contrast enhancement, geometric correction or scaling and cropping preprocessing operations on the real-time image data collected by the image real-time acquisition module.

[0014] Furthermore, the feature extraction module uses a convolutional neural network to extract key features from the preprocessed image, and then uses FasterR-CNN, YOLO algorithm or SSD target detection algorithm to perform position and category detection on the extracted target features to determine the position and category of the target.

[0015] Furthermore, the target recognition and tracking module uses a YOLO algorithm, a Kalman filter algorithm or a particle filter tracking algorithm to detect and track the target.

[0016] Furthermore, the decision module determines the driving decision through GPT model analysis based on the text information fed back by the scene understanding module.

[0017] Furthermore, the path planning module determines the vehicle's driving path according to the driving decision, using a graph search algorithm or an RRT method, and according to the type of driving road.

[0018] One or more embodiments of this specification provide an automatic driving image processing method, including the steps of:

[0019] Collect real-time image data;

[0020] The key features in the image are extracted through the feature extraction model to obtain the target features, and the extracted target features are detected to determine the position and category of each target in the image;

[0021] Further identify the detected target, determine the specific type of the target, and track the target through the target tracking algorithm;

[0022] Based on the detection and tracking results, analyze and determine traffic signals, signs, lane line information, understand traffic rules and road conditions, and generate text information;

[0023] Generate vehicle driving decisions based on the generated text information and vehicle status; and

[0024] Based on driving decisions, the vehicle's future travel path is planned and the vehicle's movement is controlled by a control unit.

[0025] Furthermore, it also includes an image preprocessing step, which performs denoising, contrast enhancement, geometric correction or scaling and cropping preprocessing operations on the real-time image data collected by the image real-time acquisition module.

[0026] Furthermore, the feature extraction module uses a convolutional neural network to extract key features from the preprocessed image, and then uses FasterR-CNN, YOLO algorithm or SSD target detection algorithm to perform position and category detection on the extracted target features to determine the position and category of the target.

[0027] One or more embodiments of the present specification provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the autonomous driving image processing method as described in any one of the above items when executing the computer program.

[0028] One or more embodiments of the present specification provide a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the autonomous driving image processing method as described in any one of the above items is implemented.

[0029] The present disclosure provides an autonomous driving image processing system, method, device and medium, which have the advantages of performing target detection based on extracted features, determining the position and category of each target in the image, further identifying the detected target, determining the specific type of the object, and then combining the detection and identification results to understand the entire scene, including traffic flow, lane line information, traffic signals, understand traffic rules and road conditions, provide context information for the decision-making module, and the decision-making module generates driving decisions based on the scene understanding and vehicle status. Finally, the vehicle's driving path is planned and driven through a path planning method based on the driving decision, ensuring safe and efficient driving, system integration and control, integrating all modules into a unified system, and ensuring coordinated work between modules. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0031] Figure 1 A block diagram of an autonomous driving image processing system provided for one or more embodiments of this specification;

[0032] Figure 2 A flowchart of an automatic driving image processing method provided for one or more embodiments of this specification;

[0033] as well as

[0034] Figure 3 A schematic diagram of the structure of a computer device provided for one or more embodiments of this specification. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0036] The present invention is described in detail below in conjunction with specific implementation methods and the accompanying drawings.

[0037] System Example

[0038] According to an embodiment of the present invention, an automatic driving image processing system is provided. Figure 1 FIG. 1 is a schematic block diagram of an automatic driving image processing system provided in this embodiment. The automatic driving image processing system according to an embodiment of the present invention includes:

[0039] The real-time image acquisition module 10 is used to acquire real-time image data through a sensor.

[0040] Target detection module 20: used to extract key features in the image through a feature extraction model, obtain target features, and detect the extracted target features to determine the position and category of each target in the image, where the targets may include vehicle contours, pedestrian postures, traffic signs, lane lines, obstacles, etc.

[0041] Target recognition and tracking module 30: used to further recognize the detected target, determine the specific type, position, speed and motion state of the target, and track the target through a target tracking algorithm.

[0042] Scene understanding module 40: used to analyze and determine traffic signals, signs, lane line information based on the detection and tracking results of the target detection module 20 and the target recognition and tracking module 30, understand traffic rules and road conditions, and generate text information to feed back to the decision module 50.

[0043] Decision module 50: used to generate vehicle driving decisions based on the text information of the scene understanding module 40 and the vehicle status, wherein the driving decisions may include acceleration, deceleration, steering, etc.

[0044] Path planning module 60: used to plan the vehicle's future driving path according to the driving decision of the decision module 50, and to control the movement of the vehicle.

[0045] In the automatic driving image processing system provided in this embodiment, the target detection module 20 performs target detection based on the extracted features to determine the position and category of each target in the image. The target recognition module 30 further recognizes the detected target to determine the specific type of the object. The scene understanding module 40 combines the detection and recognition results to understand the entire scene, including traffic flow, lane line information, traffic signals, understands traffic rules and road conditions, and provides context information for the decision-making module. According to the scene understanding and vehicle status, the decision-making module generates driving decisions. Finally, the path planning module 60 implements the planning and driving of the vehicle's driving path through a path planning method according to the driving decision, to ensure safe and efficient driving, system integration and control, and integrates all modules into a unified system to ensure coordinated work between modules.

[0046] In this embodiment, it also includes feature fusion of real-time image data collected by sensors, that is, during the autonomous driving process, the real-time image acquisition module 10 will acquire image data collected by different sensors in real time, such as point cloud of laser radar, reflection data of millimeter-wave radar and image of camera, etc., and fuse the features extracted by different sensors to obtain more comprehensive environmental information. The specific submodules are as follows:

[0047] Spatial alignment module: First, the data from different sensors need to be aligned in space. Specifically, the data collected by sensors such as the lidar point cloud, the millimeter-wave radar reflection data, and the camera image data are mapped to the same coordinate system. The alignment of the camera and the lidar is achieved through external parameter calibration, and the millimeter-wave radar reflection information can also be aligned with the image through coordinate transformation.

[0048] Feature fusion module: performs feature fusion on the data processed by the spatial alignment module. For example, the point cloud data of sensors such as radar is mapped into a depth map (depth information corresponding to the camera image), and then the feature is stacked with the image of the camera in the channel direction as the input of the target detection network.

[0049] In this embodiment, an image preprocessing module 70 is also included, which is used to perform preprocessing operations such as denoising, contrast enhancement, geometric correction, scaling and cropping functions on the real-time image data collected by the real-time image acquisition module 10. The preprocessed image data will be used as the input of the feature extraction module, wherein the image enhancement processing is to perform contrast enhancement according to the light conditions to improve the image quality in the lighting changes and shadows, so as to better identify the details in the image; geometric correction is performed to eliminate the distortion caused by vehicle movement or sensor installation angle; and noise reduction and contrast enhancement are performed on night images to improve image quality.

[0050] In this embodiment, the real-time image acquisition module 10 acquisition system integrates multiple sensor data. The real-time image acquisition module 10 is connected to each sensor through a data interface to realize real-time data transmission, including sensors such as cameras, lidars, radars and ultrasonic waves, to obtain multi-dimensional information of the vehicle's surroundings, ensuring the consistency of data collected by different sensors in time and space, so as to facilitate subsequent data processing and analysis.

[0051] In this embodiment, the feature extraction module can use a convolutional neural network to extract key features from the preprocessed image. The extracted features will be used for object detection and recognition, and then the FasterR-CNN, YOLO (You Only Look Once) or SSD (Single Shot MultiBox Detector) target detection algorithm is used to perform target detection on the extracted features to determine the target position and category in the image to adapt to the dynamic environment during high-speed driving.

[0052] The target recognition and tracking module 30 can use the YOLO algorithm, Kalman filter or particle filter tracking algorithm to detect and track the target, or it can also use the STC (Sparse and Low-Rank Rrpresentation with Temporal Consistency) target tracking algorithm, the SSD target detection algorithm tracks the target in the image data to be detected, and feeds the recognition result back to the scene understanding module 40; for example, the YOLO algorithm is used to detect the lane line position, and then the change of continuous frame rate is tracked according to the target tracking algorithm. For pedestrians in the image, and their trajectories are predicted, the model R-CNN or FasterR-CNN can be used.

[0053] In this embodiment, the scene understanding module 40 analyzes and determines the traffic signals, signs, lane line information, understands traffic rules and road conditions, and analyzes the distance, speed and relative position between vehicles, pedestrian and vehicle behaviors, and dangerous situations based on the detection and tracking results of the target detection module 20 and the target recognition and tracking module 30. For example, this module can be set to evaluate the collision risk according to the judgment rule of the collision time TTC, and then statistically calibrate the collision time TTC of different driving scenarios to obtain the collision time TTC with collision risks in different driving scenarios; specifically, first predict the trajectory of the vehicle and the trajectory of other vehicles based on the perception results of the perception model, and then predict the collision of the trajectory of the vehicle and the trajectory of other vehicles to obtain the collision time TTC of the vehicle and other vehicles, and then determine whether there is a collision risk based on the actual driving scenario and the collision time TTC.

[0054] In one example, the scene understanding module performs the following analysis on the input data:

[0055] 1) Traffic signal and sign analysis process;

[0056] First, the features of traffic signals and signs provided by the target detection module are extracted, and their color, shape, and symbol features are analyzed to determine the state of the signal (such as red light, green light) and the meaning of the sign (such as speed limit, stop, etc.). Then, combined with time series data (such as the time when the signal state changes), the current state of traffic signals and signs is updated in real time.

[0057] 2) Lane line information analysis process;

[0058] Bounding box detection: Through the lane line information provided by the object detection module, the positional relationship between the lane line and the vehicle is analyzed to determine whether there is a lane deviation and ensure that the vehicle follows the lane instructions.

[0059] 3) Vehicle distance and speed analysis process

[0060] The target recognition and tracking module obtains the location information and speed data of surrounding vehicles, calculates the distance and relative speed between vehicles in real time, and the system continuously monitors the surrounding environment, updates the relative positions between vehicles, and identifies potential collision risks.

[0061] Through the above analysis process and results, the scene understanding module can generate the following key text information:

[0062] Traffic signal status: For example, "Red light ahead" or "Speed ​​limit 40 km / h".

[0063] Lane information: For example, "driving within lane" or "departing from lane".

[0064] Status of surrounding vehicles: For example, "The vehicle ahead is 5 meters away and the speed is 40 km / h" or "A pedestrian is approaching on the left."

[0065] Finally, these text information will be output in a structured format and fed back to the decision module 50 to facilitate the decision module 50 to analyze.

[0066] In this embodiment, the decision module 50 determines the driving decision, such as turning left, turning right, going straight, slowing down, stopping, etc., through GPT model analysis based on the text information fed back by the scene understanding module 40. This embodiment utilizes the generalization and reasoning capabilities of the GPT model to solve various driving scenarios. For example, based on the feedback from the scene understanding module 40 that the green light ahead turns red, a decision to slow down and stop is generated, and the decision result will be transmitted to the path planning module 60.

[0067] In this embodiment, the path planning module 60 uses a graph search algorithm or an RRT (Rapidly-exploring Random Trees) method based on driving decisions to optimize and determine the vehicle's driving path according to the type of driving road to ensure safe and efficient autonomous driving to maintain the optimal vehicle distance and speed, wherein the driving road types include highways, urban roads, complex intersections, mountain roads, rural roads, etc.

[0068] The system provided in this embodiment can also use self-supervised learning models in the deep learning model, such as autoencoder, contrastive learning, etc., in feature extraction and object recognition, compared with traditional image processing methods, although the algorithm complexity is higher, it can achieve faster processing speed and higher accuracy, and has significant advantages. In order to reduce the model's dependence on image data, it is trained through large-scale unlabeled data (self-supervised learning) to reduce the dependence on labeled data. In addition, the system can also use real-time collected data for online learning to adaptively update the model in real time to adapt to the changing environment. For example, YOLO, Faster R-CNN) models, in real-time scenarios, these models can be fine-tuned with new image data to adapt to changes in the current environment, such as identifying new types of objects or performance under different lighting conditions.

[0069] This embodiment also provides a data management module for collecting, storing, analyzing and utilizing data in the image processing process, and a security and privacy protection module for using encryption and anonymization technology to ensure the safe operation of the system and user privacy protection, so as to protect user privacy.

[0070] The machine learning models in each module of the system of this embodiment have good versatility and are modularly designed to ensure coordinated work between modules and achieve seamless integration with other autonomous driving system components for easy maintenance and upgrading. The data management module can collect, store, analyze and utilize large-scale data sets for model training and optimization.

[0071] Method Embodiment

[0072] According to an embodiment of the present invention, a method for processing an image for autonomous driving is provided. Figure 2 As shown, it is a flow chart of the automatic driving image processing method provided in this embodiment. The automatic driving image processing method according to the embodiment of the present invention comprises the following steps:

[0073] Step S1, collecting real-time image data;

[0074] Step S2, extracting key features in the image through a feature extraction model, obtaining target features, and detecting the extracted target features to determine the position and category of each target in the image, where the targets may include vehicle contours, pedestrian postures, traffic signs, lane lines, obstacles, etc.

[0075] Step S3, further identifying the detected target, determining the specific type of the target, and tracking the target through a target tracking algorithm.

[0076] Step S4, based on the detection and tracking results, analyze and determine the traffic signals, signs, and lane line information, understand the traffic rules and road conditions, and generate text information.

[0077] Step S5, generating a vehicle driving decision based on the text information of step S4 and the vehicle status, wherein the driving decision may include acceleration, deceleration, steering, etc.

[0078] Step S6, planning the vehicle's future travel path based on the driving decision, and controlling the vehicle's movement through the vehicle's control unit.

[0079] The method provided in this embodiment performs target detection based on the extracted features, determines the position and category of each target in the image, further identifies the detected target, determines the specific type of the object, and then combines the detection and identification results to understand the entire scene, including traffic flow, lane line information, traffic signals, understand traffic rules and road conditions, provide context information for the decision-making module, and then combine the vehicle status to generate a driving decision. Finally, according to the driving decision, the path planning method is used to plan and drive the vehicle's driving path to ensure safe and efficient driving.

[0080] In this embodiment, an image preprocessing step is also included, in which the collected real-time image data is subjected to preprocessing operations such as denoising, contrast enhancement, geometric correction, scaling and cropping functions. The preprocessed image data will be used as the input of the feature extraction module, wherein the image enhancement processing is to perform contrast enhancement according to the light conditions to improve the image quality in the light changes and shadows so as to better identify the details in the image; geometric correction is performed to eliminate the distortion caused by the vehicle movement or the sensor installation angle; and noise reduction and contrast enhancement processing are performed on the night image to improve the image quality.

[0081] In step S3 of this embodiment, the feature extraction module can use a convolutional neural network to extract key features from the preprocessed image. The extracted features will be used for object detection and recognition, and then the FasterR-CNN, YOLO (You Only LookOnce, target detection algorithm) or SSD (Single Shot MultiBox Detector) target detection algorithm is used to detect the extracted target features and determine the location and category of the target to adapt to the dynamic environment during high-speed driving.

[0082] In step S3 of this embodiment, the YOLO algorithm, Kalman filter or particle filter tracking algorithm may be used to detect and track the target, or the STC (Sparse and Low-Rank Rrpresentation with Temporal Consistency) target tracking algorithm, SSD target detection algorithm, etc. may be used to track the target in the image data to be detected, and the recognition result is fed back to the scene understanding module 40;

[0083] In step S5 of this embodiment, the driving decision is determined through GPT model analysis based on the text information generated in step S4.

[0084] In step S6 of this embodiment, based on the driving decision, a graph search algorithm or an RRT method is used to optimize the vehicle's driving path according to the type of road to ensure safe and efficient autonomous driving to maintain the optimal vehicle distance and speed.

[0085] The embodiment of the present invention is a method embodiment corresponding to the above-mentioned system embodiment. The specific operation of each step can be understood by referring to the module processing description of the system embodiment, which will not be repeated here.

[0086] like Figure 3 As shown, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the autonomous driving image processing method in the above-mentioned embodiment is implemented.

[0087] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the autonomous driving image processing method in the above-mentioned embodiment is implemented; or when the computer program is executed by a processor, the autonomous driving image processing method in the above-mentioned embodiment is implemented.

[0088] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0089] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and the contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. An automatic driving image processing system, characterized in that include: The real-time image acquisition module is used to acquire real-time image data through sensors; Target detection module: used to extract key features in the image through the feature extraction model, obtain target features, and detect the extracted target features to determine the location and category of each target in the image; Target recognition and tracking module: used to further identify the detected target, determine the specific type of the target, and track the target through the target tracking algorithm; Scene understanding module: used to analyze and determine traffic signals, signs, lane line information, understand traffic rules and road conditions, and generate text information to feed back to the decision module based on the detection and tracking results of the target detection module and the target recognition and tracking module; Decision module: used to generate vehicle driving decisions based on the text information and vehicle status of the scene understanding module; as well as Path planning module: used to plan the vehicle's future driving path based on the driving decisions of the decision-making module, and to control the movement of the vehicle.

2. The automatic driving image processing system according to claim 1, characterized in that: An image preprocessing module is also provided to perform denoising, contrast enhancement, geometric correction or scaling and cropping preprocessing operations on the real-time image data collected by the image real-time acquisition module.

3. The automatic driving image processing system according to claim 1, characterized in that: The feature extraction module uses a convolutional neural network to extract key features from the preprocessed image, and then uses FasterR-CNN, YOLO algorithm or SSD target detection algorithm to perform position and category detection on the extracted target features to determine the position and category of the target.

4. The automatic driving image processing system according to claim 1, characterized in that: The target recognition and tracking module uses the YOLO algorithm, Kalman filter algorithm or particle filter tracking algorithm to detect and track the target.

5. The automatic driving image processing system according to claim 1, characterized in that: The decision module determines the driving decision through GPT model analysis based on the text information fed back by the scene understanding module; The path planning module determines the vehicle's driving path based on the driving decision, using a graph search algorithm or RRT method, depending on the type of road being driven.

6. An automatic driving image processing method, characterized in that Includes steps: Collect real-time image data; The key features in the image are extracted through the feature extraction model to obtain the target features, and the extracted target features are detected to determine the position and category of each target in the image; Further identify the detected target, determine the specific type of the target, and track the target through the target tracking algorithm; Based on the detection and tracking results, analyze and determine traffic signals, signs, lane line information, understand traffic rules and road conditions, and generate text information; Generate vehicle driving decisions based on the generated text information and vehicle status; as well as Based on driving decisions, the vehicle's future travel path is planned and the vehicle's movement is controlled by a control unit.

7. The automatic driving image processing method according to claim 6, characterized in that: The method also includes an image preprocessing step, which performs denoising, contrast enhancement, geometric correction or scaling and cropping preprocessing operations on the real-time image data collected by the image real-time acquisition module.

8. The automatic driving image processing method according to claim 6, characterized in that: The feature extraction module uses a convolutional neural network to extract key features from the preprocessed image, and then uses FasterR-CNN, YOLO algorithm or SSD target detection algorithm to perform position and category detection on the extracted target features to determine the position and category of the target.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the automatic driving image processing method as described in any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the automatic driving image processing method according to any one of claims 1 to 5 is implemented.

Images