Recognition and obstacle avoidance method and system for super-long truck in automatic driving system

Abstract

The invention discloses a recognition and obstacle avoidance method and system for a super-long truck in an automatic driving system, and the method comprises the following steps: a data collection step: employing a multi-sensor fusion technology, and collecting the point cloud data, image data and radar data of the surrounding environment of a vehicle; an image processing step: preprocessing the collected data; a segmentation identification step: carrying out segmentation identification on the super-long truck, and extracting geometric features and texture features of the truck part; a vehicle matching step: matching the segmented and identified vehicle parts through a target association algorithm; a vehicle splicing step: splicing the segmented and identified vehicle parts by using a geometric splicing algorithm and a motion consistency algorithm; and a decision and control step: outputting the contour, the position and the motion state of the complete vehicle, and carrying out decision and control on the vehicle. According to the method, a segmented identification and splicing algorithm is adopted, the identification problem caused by an overlong truck body of an overlong truck is effectively solved, and the identification efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and particularly to a method and system for identifying and avoiding obstacles of extra-long trucks in an autonomous driving system. Background Art

[0002] With the rapid development of autonomous driving technology, significant progress has been made in core modules such as environmental perception, decision-making and planning, and control execution. The advancement of sensor technologies (such as lidar, millimeter-wave radar, cameras) enables an autonomous driving system to perceive the surrounding environment in real time and identify targets such as vehicles, pedestrians, and obstacles. At the same time, the application of deep learning algorithms has greatly improved the accuracy of target detection, tracking, and classification. In addition, the introduction of high-precision maps and vehicle networking technologies provides the autonomous driving system with global environmental information and collaborative perception capabilities.

[0003] In the freight transportation field, extra-long trucks (such as trailers, container trucks) have become an important part of road traffic due to their special body structures and driving characteristics. However, the length of extra-long trucks far exceeds that of ordinary vehicles, posing additional challenges to the environmental perception and obstacle avoidance of autonomous driving systems. Traditional autonomous driving systems are often optimized for vehicles of standard length, and for extra-long trucks, their detection, identification, and obstacle avoidance strategies may not be precise and efficient enough. During actual driving, this may lead to the autonomous driving system being unable to timely identify the presence of an extra-long truck or accurately judge its driving trajectory and speed, thus increasing the risk of collision. Summary of the Invention

[0004] The technical problem solved by the present invention is to provide a method for identifying and avoiding obstacles of extra-long trucks in an autonomous driving system that can achieve obstacle avoidance for extra-long trucks.

[0005] The technical solution adopted by the present invention to solve its technical problem is: A method for identifying and avoiding obstacles of extra-long trucks in an autonomous driving system, comprising the following steps:

[0006] Data acquisition step: Using multi-sensor fusion technology, collect point cloud data, image data, and radar data of the vehicle surrounding environment;

[0007] Image processing step: Preprocess the collected data, including point cloud noise reduction and segmentation, image enhancement, and target detection;

[0008] Segmented identification step: Segment and identify an extra-long truck, and extract geometric features and texture features of the vehicle part;

[0009] Vehicle matching step: Match the segmented vehicle parts through a target association algorithm;

[0010] Vehicle splicing steps: Use geometric splicing algorithms and motion consistency algorithms to splice the segmented vehicle parts into a complete vehicle;

[0011] Decision and control steps: Output the outline, position, and motion state of the complete vehicle, and perform vehicle decision-making and control through autonomous driving algorithms.

[0012] Furthermore: In the data acquisition step, specifically:

[0013] Use lidar to collect high-precision 3D point cloud data;

[0014] Use millimeter-wave radar for long-distance target detection and speed measurement;

[0015] Use cameras for target recognition and feature extraction.

[0016] Furthermore: In the image processing step, specifically:

[0017] Perform noise reduction and segmentation on the point cloud data to remove outliers and ground point clouds;

[0018] Perform denoising, dehazing, and HDR processing on the image data;

[0019] Use a deep learning model to perform target detection on the image data and extract the texture features of the vehicle parts.

[0020] Furthermore: In the segmentation recognition step, specifically:

[0021] Use a deep learning model to extract the geometric features and texture features of the vehicle parts;

[0022] Extract the geometric information of length, width, and height from the point cloud data;

[0023] Extract the logo information of the vehicle from the image data;

[0024] The vehicle matching step, specifically:

[0025] Use the Hungarian algorithm or Kalman filter to associate the targets detected by different sensors;

[0026] Perform matching according to the motion state and geometric features of the targets.

[0027] Furthermore: The geometric splicing algorithm is: Use the least squares method or the RANSAC algorithm to register the point cloud data of the vehicle parts, and splice them into a complete vehicle according to the geometric features of the vehicle parts;

[0028] The motion consistency algorithm is: Verify the consistency of the splicing result according to the motion state of the vehicle parts, and use the motion model for correction.

[0029] Furthermore: in the decision-making and control step, the outline, position, and motion state of the complete vehicle are output, and the decision-making and control of the vehicle are carried out through the autonomous driving algorithm, and a safe distance from the ultra-long truck is maintained. Specifically:

[0030] Obtain the length, speed, and acceleration of the ultra-long truck, and the speed and acceleration of the host vehicle;

[0031] Obtain the road condition information, and obtain the time distance t according to the road condition information;

[0032] Obtain the safe distance D:

[0033]

[0034] where v is the speed of the host vehicle, t is the time distance, a is the acceleration of the host vehicle, and L is the length of the ultra-long truck.

[0035] Furthermore: it also includes a multi-level obstacle avoidance strategy, specifically including:

[0036] First-level obstacle avoidance: Start the deceleration strategy and use the smooth control algorithm to adjust the vehicle speed;

[0037] Second-level obstacle avoidance: Start the lane-changing strategy and use the path planning algorithm to generate a lane-changing trajectory;

[0038] Third-level obstacle avoidance: Start the emergency stop strategy and use the maximum braking force to stop the vehicle.

[0039] The present invention also discloses an identification and obstacle avoidance system for ultra-long trucks in an autonomous driving system, including a data acquisition module, an image processing module, a segment identification module, a vehicle matching module, a vehicle splicing module, and a decision-making and control module;

[0040] The data acquisition module is used to collect the point cloud data, image data, and radar data of the vehicle surrounding environment through multi-sensor fusion technology;

[0041] The image processing module is used to preprocess the collected data, including point cloud noise reduction and segmentation, image enhancement, and target detection;

[0042] The segment identification module is used to segment and identify the ultra-long truck, and extract the geometric features and texture features of the vehicle part;

[0043] The vehicle matching module is used to match the segmented vehicle parts through the target association algorithm;

[0044] The vehicle splicing module is used to splice the segmented vehicle parts into a complete vehicle through the geometric splicing algorithm and the motion consistency algorithm;

[0045] The decision-making and control module is used to output the outline, position, and motion state of the complete vehicle, and perform vehicle decision-making and control through an autonomous driving algorithm.

[0046] The present invention also discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for identifying and avoiding obstacles of an extra-long truck in the above-mentioned autonomous driving system are realized.

[0047] The present invention also discloses a computer device, including a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete communication with each other through the communication bus; where:

[0048] The memory is used to store a computer program;

[0049] The processor is used to execute the steps of the method for identifying and avoiding obstacles of an extra-long truck in the above-mentioned autonomous driving system by running the program stored on the memory.

[0050] The beneficial effects of the present invention are:

[0051] 1. Through multi-sensor fusion technology (lidar, millimeter-wave radar, camera), the system can accurately capture the complete outline and dynamic behavior of an extra-long truck, significantly improving the recognition accuracy.

[0052] 2. By adopting a segmented recognition and splicing algorithm, the recognition problem caused by the too long body of an extra-long truck is effectively solved, and the recognition efficiency is improved.

[0053] 3. The formulation of a multi-level obstacle avoidance strategy ensures that the autonomous driving vehicle can take timely safety measures when encountering an extra-long truck, greatly reducing the collision risk.

[0054] 4. The concept of "Time Gap" is introduced, and the safety distance is dynamically adjusted according to the length, speed, and road conditions of the extra-long truck to ensure a safe interval in different scenarios.

[0055] 5. The system for identifying and avoiding obstacles of an extra-long truck in the autonomous driving system of the present invention has a clear structure, reasonable module division, and is easy to implement and maintain. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic flowchart of the method for identifying and avoiding obstacles of an extra-long truck in the autonomous driving system according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] In order to make the above-mentioned objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention is made in conjunction with the accompanying drawings.

[0058] As Figure 1 shown, an embodiment of the present application discloses a method for identifying and avoiding obstacles for an extra-long truck in an autonomous driving system, including the following steps:

[0059] Data acquisition step: Using multi-sensor fusion technology, acquire point cloud data, image data, and radar data of the vehicle surrounding environment;

[0060] Image processing step: Preprocess the acquired data, including point cloud noise reduction and segmentation, image enhancement, and target detection;

[0061] Segmented identification step: Segment and identify the extra-long truck, and extract geometric features and texture features of the vehicle part;

[0062] Vehicle matching step: Match the segmented vehicle parts through a target association algorithm;

[0063] Vehicle splicing step: Use a geometric splicing algorithm and a motion consistency algorithm to splice the segmented vehicle parts into a complete vehicle;

[0064] Decision and control step: Output the contour, position, and motion state of the complete vehicle, and perform vehicle decision and control through an autonomous driving algorithm.

[0065] In this method, through multi-sensor fusion technology and techniques such as segmental acquisition of the vehicle, more accurate identification results of the extra-long truck can be obtained. Finally, in the decision and control step, the system outputs the contour, position, and motion state of the complete vehicle, and performs vehicle decision and control through an autonomous driving algorithm, thereby greatly reducing the collision risk with the extra-long truck.

[0066] In this embodiment, in the data acquisition step, specifically:

[0067] Use a lidar to acquire high-precision 3D point cloud data;

[0068] Use a millimeter-wave radar for long-distance target detection and speed measurement;

[0069] Use a camera for target recognition and feature extraction.

[0070] Specifically, the lidar emits laser beams and receives the reflected signals to generate 3D point cloud data of the vehicle surrounding environment. These data provide accurate position and shape information of the objects around the vehicle. The millimeter-wave radar uses high-frequency electromagnetic waves to detect distant targets and can accurately measure the distance and speed of the targets, providing real-time dynamic information for the system. The camera captures images around the vehicle and performs target recognition and feature extraction through image processing algorithms, such as the shape, color, license plate number, etc. of the vehicle, providing rich visual information for the system.

[0071] In the above steps, through multi-sensor fusion technology (lidar, millimeter-wave radar, camera), the system can accurately capture the complete contour and dynamic behavior of the ultra-long truck, significantly improving the recognition accuracy.

[0072] In this embodiment, the image processing step is specifically as follows:

[0073] Perform noise reduction and segmentation on the point cloud data to remove outliers and ground point clouds;

[0074] Perform denoising, defogging, and HDR processing on the image data;

[0075] Use a deep learning model to perform object detection on the image data and extract the texture features of the vehicle part.

[0076] Specifically, performing noise reduction processing on the point cloud data can remove abnormal points caused by sensor noise or environmental factors, improving the accuracy of the data. Point cloud segmentation divides the point cloud data into different parts, such as the ground, vehicles, buildings, etc., for subsequent processing and analysis. Performing denoising processing on the image data can remove noise points in the image and improve the clarity of the image. Defogging processing can remove the fog in the image and enhance the contrast of the image. HDR (High Dynamic Range) processing can expand the dynamic range of the image, making the bright and dark details in the image clearer. Using a deep learning model to perform object detection on the image data can automatically identify the vehicles in the image and extract the texture features of the vehicles, such as license plate numbers, vehicle colors, etc., providing key information for subsequent vehicle recognition and matching.

[0077] In the above steps, through the image processing step, the system can further process and analyze the collected data, extract the key features of the ultra-long truck, and provide a basis for subsequent vehicle recognition and obstacle avoidance decision-making.

[0078] In this embodiment, the segmented recognition step is specifically as follows:

[0079] Use a deep learning model to extract the geometric features and texture features of the vehicle part;

[0080] Extract the geometric information of length, width, and height from the point cloud data;

[0081] Extract the logo information of the vehicle from the image data;

[0082] The vehicle matching step is specifically as follows:

[0083] Use the Hungarian algorithm or Kalman filter to associate the targets detected by different sensors;

[0084] Match according to the motion state and geometric features of the target.

[0085] Specifically, in the segmented recognition step, a deep learning model is used to extract the geometric features and texture features of the vehicle parts. These features are crucial for the subsequent vehicle splicing and decision-making and control steps. The geometric information such as length, width, and height extracted from the point cloud data helps the system accurately understand the physical size and space occupancy of the extra-long truck. At the same time, the vehicle logo information extracted from the image data, such as license plate number, vehicle type features, etc., provides important clues for the confirmation and matching of the vehicle identity.

[0086] The vehicle matching step associates the targets detected by different sensors through techniques such as the Hungarian algorithm or Kalman filtering. These algorithms can match the segmented vehicle parts according to the motion state and geometric features of the target, so as to ensure that the system can accurately identify the complete extra-long truck. This process is crucial for the subsequent vehicle splicing and obstacle avoidance decision-making, because it directly relates to whether the system can correctly understand and respond to the presence of the extra-long truck.

[0087] Through the above steps, the system can accurately capture the complete contour and dynamic behavior of the extra-long truck, and in the decision-making and control step, according to the length, speed, acceleration of the extra-long truck and road conditions information, dynamically adjust the safety distance to ensure a safe interval between the autonomous vehicle and the extra-long truck.

[0088] In this embodiment, the geometric splicing algorithm is: using the least squares method or the RANSAC algorithm to register the point cloud data of the vehicle parts, and splicing them into a complete vehicle according to the geometric features of the vehicle parts;

[0089] The motion consistency algorithm is: verifying the consistency of the splicing result according to the motion state of the vehicle parts, and using the motion model for correction.

[0090] Specifically, the geometric splicing algorithm uses the least squares method or the RANSAC algorithm to register the point cloud data of the segmented vehicle parts. The least squares method finds the best function match for the data by minimizing the sum of the squared errors, while the RANSAC algorithm is an iterative method used to estimate the parameter model from a dataset containing a large number of outliers. Both of these algorithms can effectively handle the noise and outliers in the point cloud data, thereby improving the accuracy of splicing.

[0091] The motion consistency algorithm verifies the consistency of the splicing result according to the motion state of the vehicle parts. The algorithm judges whether the splicing result conforms to the actual motion of the vehicle by comparing the motion trajectories and speeds of the vehicle parts in different time periods. If there are inconsistent situations, the algorithm will use the motion model to correct the splicing result to ensure the accuracy and reliability of the splicing result.

[0092] By combining the above geometric stitching algorithm and motion consistency algorithm, the system can accurately stitch the segmented vehicle parts into a complete vehicle, thereby providing accurate vehicle information for subsequent decision-making and control steps. The implementation of this step not only improves the accuracy and efficiency of over-length truck recognition, but also provides strong guarantee for the safe driving of autonomous vehicles.

[0093] In this embodiment, the decision-making and control steps: output the outline, position and motion state of the complete vehicle, make decisions and control the vehicle through the autonomous driving algorithm, and maintain a safe distance from the over-length truck. Specifically:

[0094] Obtain the length, speed, acceleration of the over-length truck, and the speed and acceleration of the host vehicle;

[0095] Obtain road condition information and obtain the time distance t according to the road condition information;

[0096] Obtain the safe distance D:

[0097]

[0098] Wherein, v is the speed of the host vehicle, t is the time distance, a is the acceleration of the host vehicle, and L is the length of the over-length truck.

[0099] Specifically, in this structure, by setting the safe distance D, the distance from the over-length truck can be adjusted flexibly. In addition, the time distance can be set according to the road conditions. The higher the vehicle speed, the greater the time distance. The more complex the road conditions, the greater the time distance. For over-length trucks, due to their longer braking distance, an additional time distance needs to be increased. For example: when driving on the highway, the speed of the host vehicle is 100 km / h (27.78 m / s), the vehicle in front is an over-length truck with a length L = 20 m, the road surface is dry, and the visibility is good. Since the vehicle speed is high and the vehicle in front is an over-length truck, set t = 3 seconds. At this time

[0100] In this method, by dynamically adjusting the time distance, the autonomous driving system can maintain a reasonable safe distance in different scenarios, effectively reducing the accident risk and improving the driving safety and comfort.

[0101] In this embodiment, it also includes a multi-level obstacle avoidance strategy, specifically including:

[0102] First-level obstacle avoidance: Start the deceleration strategy and use the smooth control algorithm to adjust the vehicle speed;

[0103] Second-level obstacle avoidance: Start the lane-changing strategy and use the path planning algorithm to generate a lane-changing trajectory;

[0104] Third-level obstacle avoidance: Start the emergency stop strategy and use the maximum braking force to stop the vehicle.

[0105] Specifically, in the first-level obstacle avoidance strategy, when the system detects that the distance from an extra-long truck is less than the preset safe distance, it will immediately activate the deceleration strategy. This strategy uses a smooth control algorithm to smoothly adjust the vehicle speed, ensuring that the vehicle gradually reduces its speed without causing sudden braking or abrupt movements, so as to maintain a safe distance from the extra-long truck. The application of the smooth control algorithm not only improves the driving smoothness but also effectively avoids rear-end collisions that may be caused by sudden braking.

[0106] In the second-level obstacle avoidance strategy, if the deceleration strategy cannot effectively avoid the potential collision risk with the extra-long truck, the system will activate the lane-changing strategy. This strategy uses a path planning algorithm to analyze the current road conditions and the driving conditions of surrounding vehicles in real time, and intelligently plan a safe lane-changing path. The path planning algorithm comprehensively considers various factors such as road curvature, traffic signals, and the driving trajectories of other vehicles to ensure the safety and efficiency of the lane-changing process. During the lane-changing process, the system will also continuously monitor the distances from the extra-long truck and other vehicles to ensure the smooth completion of the lane-changing action.

[0107] In the third-level obstacle avoidance strategy, as the final safety guarantee measure, when the system determines that the collision risk is extremely high and cannot be avoided by the deceleration or lane-changing strategies, it will immediately activate the emergency stop strategy. This strategy enables the maximum braking force of the vehicle to make the vehicle stop smoothly in the shortest time, thereby minimizing the collision risk. Although the application of the emergency stop strategy may cause the interruption of vehicle driving, it plays a crucial role in ensuring driving safety.

[0108] The present invention also discloses an identification and obstacle avoidance system for extra-long trucks in an autonomous driving system, which is characterized by including a data acquisition module, an image processing module, a segmented identification module, a vehicle matching module, a vehicle splicing module, and a decision-making and control module;

[0109] The data acquisition module is used to collect point cloud data, image data, and radar data of the vehicle surrounding environment through multi-sensor fusion technology;

[0110] The image processing module is used to preprocess the collected data, including point cloud noise reduction and segmentation, image enhancement, and target detection;

[0111] The segmented identification module is used to segment and identify the extra-long truck, and extract the geometric features and texture features of the vehicle parts;

[0112] The vehicle matching module is used to match the segmented vehicle parts through a target association algorithm;

[0113] The vehicle splicing module is used to splice the segmented recognized vehicle parts into a complete vehicle through geometric splicing algorithms and motion consistency algorithms;

[0114] The decision-making and control module is used to output the outline, position, and motion state of the complete vehicle, and perform vehicle decision-making and control through autonomous driving algorithms.

[0115] This system can realize the recognition and obstacle avoidance of extra-long trucks in the autonomous driving system, thereby ensuring driving safety.

[0116] The present invention also discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for recognizing and avoiding obstacles of extra-long trucks in the above-mentioned autonomous driving system are realized.

[0117] The present invention also discloses a computer device, including a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete mutual communication through the communication bus; where:

[0118] The memory is used to store a computer program;

[0119] The processor is used to execute the steps of the method for recognizing and avoiding obstacles of extra-long trucks in the above-mentioned autonomous driving system by running the program stored on the memory.

[0120] The above specific embodiments have further elaborated on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for identifying and avoiding obstacles of an ultra-long truck in an autonomous driving system, characterized in that, It includes the following steps: Data acquisition step: Using multi-sensor fusion technology, acquire point cloud data, image data, and radar data of the vehicle surrounding environment; Image processing step: Preprocess the acquired data, including point cloud denoising and segmentation, image enhancement and target detection; Segmented recognition step: Perform segmented recognition on the extra-long truck, and extract geometric features and texture features of the vehicle parts; Vehicle matching step: Match the vehicle parts recognized by segmentation through a target association algorithm; Vehicle splicing step: Use geometric splicing algorithm and motion consistency algorithm to splice the vehicle parts recognized by segmentation into a complete vehicle; Decision and control step: Output the contour, position, and motion state of the complete vehicle, and perform decision and control of the vehicle through an autonomous driving algorithm.

2. The method for identifying and avoiding obstacles of an extra-long truck in the automatic driving system according to claim 1, wherein, In the data acquisition step, specifically: Use lidar to acquire high-precision 3D point cloud data; Use millimeter-wave radar for long-distance target detection and speed measurement; Use a camera for target recognition and feature extraction.

3. The method for identifying and avoiding obstacles of an extra-long truck in the automatic driving system according to claim 2, wherein, The image processing step, specifically: Perform denoising and segmentation on the point cloud data, and remove outlier points and ground point clouds; Perform denoising, defogging, and HDR processing on the image data; Use a deep learning model to perform target detection on the image data and extract texture features of the vehicle parts.

4. The method for identifying and avoiding obstacles of an extra-long truck in the automatic driving system according to claim 2, characterized in that, The segmented recognition step, specifically: Use a deep learning model to extract geometric features and texture features of the vehicle parts; Extract geometric information of length, width, and height from the point cloud data; Extract logo information of the vehicle from the image data. The vehicle matching step, specifically: Use the Hungarian algorithm or Kalman filter to associate the targets detected by different sensors; Perform matching according to the motion state and geometric features of the targets.

5. The method for identifying and avoiding obstacles of an extra-long truck in the automatic driving system according to claim 1, wherein, The geometric splicing algorithm is: Use the least squares method or RANSAC algorithm to register the point cloud data of the vehicle parts, and splice them into a complete vehicle according to the geometric features of the vehicle parts; The motion consistency algorithm is: Verify the consistency of the splicing result according to the motion state of the vehicle parts, and use a motion model for correction.

6. The method for identifying and avoiding obstacles of an extra-long truck in the automatic driving system according to claim 1, characterized in that, The decision and control step: Output the contour, position, and motion state of the complete vehicle, perform decision and control of the vehicle through an autonomous driving algorithm, and maintain a safe distance from the extra-long truck. Specifically: Obtain the length, speed, and acceleration of the extra-long truck, and the speed and acceleration of the host vehicle; Obtain road condition information, and obtain the time distance t according to the road condition information; Obtain the safe distance D: Where, v is the speed of the host vehicle, t is the time distance, a is the acceleration of the host vehicle, and L is the length of the extra-long truck.

7. The method for identifying and avoiding obstacles of an extra-long truck in the automatic driving system according to claim 6, characterized in that, It also includes a multi-level obstacle avoidance strategy, specifically including: First-level obstacle avoidance: Activate the deceleration strategy and use a smooth control algorithm to adjust the vehicle speed; Second-level obstacle avoidance: Activate the lane-changing strategy and use a path planning algorithm to generate a lane-changing trajectory; Third-level obstacle avoidance: Activate the emergency stop strategy and use the maximum braking force to stop the vehicle.

8. An identification and obstacle avoidance system for an extra-long truck in an autonomous driving system, characterized in that, It includes a data acquisition module, an image processing module, a segmented recognition module, a vehicle matching module, a vehicle splicing module, and a decision and control module; The data acquisition module is used to acquire point cloud data, image data, and radar data of the vehicle surrounding environment through multi-sensor fusion technology; The image processing module is used to preprocess the collected data, including point cloud noise reduction and segmentation, image enhancement and target detection; The segment recognition module is used to recognize the extra-long truck in segments, and extract the geometric features and texture features of the vehicle parts; The vehicle matching module is used to match the vehicle parts recognized in segments through a target association algorithm; The vehicle splicing module is used to splice the vehicle parts recognized in segments into a complete vehicle through a geometric splicing algorithm and a motion consistency algorithm; The decision-making and control module is used to output the contour, position and motion state of the complete vehicle, and make decisions and control the vehicle through an automatic driving algorithm.

9. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method for recognizing and avoiding obstacles of an extra-long truck in the automatic driving system according to any one of claims 1-7 are implemented.

10. A computer device, characterized in that, It includes a processor, a communication interface, a memory and a communication bus. Among them, the processor, the communication interface and the memory complete mutual communication through the communication bus; where: The memory is used to store a computer program; The processor is used to execute the steps of the method for recognizing and avoiding obstacles of an extra-long truck in the automatic driving system according to any one of claims 1-7 by running the program stored on the memory.

Images