Obstacle avoidance control method, system, vehicle and storage medium for autonomous driving vehicle

Abstract

The present invention discloses an obstacle avoidance control method, system, vehicle and storage medium for an autonomous driving vehicle, and relates to the field of autonomous driving technology. The control method comprises: obtaining camera detection data and laser radar detection data on the autonomous driving vehicle, wherein the camera detection data comprises obstacle image information, and the laser radar detection data comprises obstacle distance information; processing the camera detection data by an image enhancement algorithm to obtain an obstacle area image, and processing the laser radar detection data by a point cloud clustering algorithm to obtain obstacle detection information; fusing the obstacle area image and the obstacle detection information to obtain multimodal information fusion data; and generating autonomous driving vehicle anti-collision warning information according to the multimodal information fusion data and a preset autonomous driving vehicle anti-collision warning system. The present invention effectively improves the accuracy of detection by simultaneously utilizing the combination of camera data and laser radar data for vehicle anti-collision detection.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving vehicles, and in particular to an autonomous driving vehicle obstacle avoidance control method, system, vehicle and storage medium. Background Art

[0002] At present, the active obstacle avoidance system of autonomous driving vehicles mainly collects information about the current driving environment and performs risk analysis through the environmental perception system. When an emergency occurs, the vehicle is actively turned to avoid obstacles. Therefore, reasonable obstacle avoidance decisions are of great significance to improving the safety of autonomous driving.

[0003] However, self-driving car collisions also occur from time to time. In order to reduce the number of self-driving car collisions, existing technologies generally use camera sensors or lidar for detection. However, camera sensors are easily affected by the environment, while lidar is not affected by the environment but is expensive.

[0004] In the related technology, the algorithm for designing an active obstacle avoidance system has greatly increased the complexity and computational complexity of the algorithm, thereby increasing the data processing time. As a result, the current autonomous driving vehicles have low safety when performing obstacle avoidance actions, and it is difficult to meet the real-time requirements of autonomous driving vehicle collision avoidance application scenarios. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide an obstacle avoidance control method, system, vehicle and storage medium for an autonomous driving vehicle, so as to solve the technical problems that the existing autonomous driving system has large algorithm calculation amount in obstacle avoidance detection and the real-time and accuracy of anti-collision detection data need to be improved.

[0006] In order to solve the above problems, the first object of the present invention is to provide an obstacle avoidance control method for an autonomous driving vehicle, comprising the following steps:

[0007] Step S 100 : Obtaining camera detection data and laser radar detection data on the autonomous driving vehicle, wherein the camera detection data includes obstacle image information, and the laser radar detection data includes obstacle distance information;

[0008] Step S 200 : Processing the camera detection data through an image enhancement algorithm to obtain an obstacle area image, and processing the laser radar detection data through a point cloud clustering algorithm to obtain obstacle detection information;

[0009] Step S 300 : Fusing the obstacle area image and the obstacle detection information to obtain multimodal information fusion data;

[0010] Step S 400: Generate autonomous driving vehicle anti-collision warning information based on the multimodal information fusion data and a preset autonomous driving vehicle anti-collision warning system.

[0011] Furthermore, in step S 200 In the method, the camera detection data is processed by an image enhancement algorithm to obtain an obstacle area image, comprising the following steps:

[0012] Acquire image data of the surrounding environment through the vehicle-mounted camera;

[0013] Preprocessing the image data, wherein the preprocessing includes denoising, enhancement, and transformation operations;

[0014] Combining the logarithmic image processing model and Lee image enhancement algorithm, the preprocessed image data is enhanced to obtain balanced image data;

[0015] Extracting features related to obstacle boundaries from the equalized image data;

[0016] Use deep learning algorithms to classify and identify the extracted features to determine the type and location of obstacles;

[0017] The obstacle image is filtered by a feature filter to obtain an obstacle area image.

[0018] Furthermore, the calculation expression of the Lee image enhancement algorithm is:

[0019]

[0020] in: Represents the pixel brightness value of the processed image, is the pixel brightness value of the original image, So centered The arithmetic mean of the window, and is a real number, [•] + Indicates taking a positive value, that is, if the value in the brackets is negative, take 0; if it is positive, take itself.

[0021] Furthermore, in step S 200 In the method, the laser radar detection data is processed by a point cloud clustering algorithm to obtain obstacle detection information, including:

[0022] Determine the size of the two-dimensional matrix based on the horizontal and vertical viewing angles of the laser radar, determine the coordinates of each corresponding point cloud in the two-dimensional matrix based on the distance and orientation of each laser point relative to the laser radar in the horizontal and vertical directions, and perform matrix filling on the two-dimensional matrix based on the three-dimensional coordinates of each point and its distance relative to the laser radar;

[0023] Calculate the elevation angle and elevation angle difference in the vertical direction for adjacent laser points, and mark the points where the elevation angle and elevation angle difference are both less than the corresponding threshold as ground points;

[0024] Calculate the point cloud distance for adjacent laser points;

[0025] The curvature in the horizontal direction at the location of each laser point is calculated using the point cloud polar coordinate curvature calculation formula, and the result is stored as a point cloud curvature two-dimensional matrix based on the coordinates of the point cloud in the two-dimensional matrix;

[0026] Combining the point cloud curvature two-dimensional matrix with the corresponding curvature threshold to determine whether they belong to the same obstacle category;

[0027] Based on the determined category results, breadth-first search clustering is performed on all points, and the same number is set for the same type of obstacle points and stored as a two-dimensional matrix of clustering results, thereby realizing the identification of different obstacles.

[0028] Furthermore, the point cloud polar coordinate curvature calculation formula is:

[0029]

[0030] in, is the horizontal angle of a point relative to the laser radar, is the elevation angle of a point relative to the laser radar, is the polar diameter based on the polar coordinate system, , Represents the polar radius of n neighboring points.

[0031] Furthermore, combining the point cloud curvature two-dimensional matrix with the corresponding curvature threshold to determine whether the objects belong to the same obstacle category includes:

[0032] Calculate the deviation of the curvature of each laser point in the horizontal direction from the center of cluster curvature;

[0033] The laser radar point cloud data to be clustered whose deviation is less than or equal to the corresponding curvature threshold is used as new laser point cloud detection data to be clustered.

[0034] Furthermore, the step of fusing the obstacle area image and the obstacle detection information to obtain multimodal information fusion data includes:

[0035] Step S 310: Performing data alignment on the obstacle area image and the obstacle detection information;

[0036] Step S 320 : Associating the aligned obstacle area image with the obstacle detection information, wherein the association process is used to match the same obstacles;

[0037] Step S 330 : Generate environmental perception information as the multimodal information fusion data based on the association result, the three-dimensional object position information in the obstacle area image, and the obstacle distance, obstacle movement speed and obstacle movement angle of the obstacle detection information.

[0038] A second object of the present invention is to provide an obstacle avoidance system for an autonomous driving vehicle, comprising:

[0039] An acquisition module, used to acquire camera detection data and laser radar detection data, wherein the camera detection data includes obstacle image information and the laser radar detection data includes obstacle distance information;

[0040] An image processing module is used to process the camera detection data through an image enhancement algorithm to obtain an obstacle area image;

[0041] A laser radar processing module, used to process the laser radar detection data through a point cloud clustering algorithm to obtain obstacle detection information;

[0042] A data fusion module, used for fusing the obstacle area image and the obstacle detection information to obtain multi-modal information fusion data;

[0043] The warning module is used to generate the anti-collision warning information of the autonomous driving vehicle according to the multimodal information fusion data and the preset anti-collision warning system of the autonomous driving vehicle.

[0044] The third object of the present invention is to provide an autonomous driving vehicle, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the autonomous driving vehicle obstacle avoidance control method as described above.

[0045] The fourth object of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the obstacle avoidance control method for an autonomous driving vehicle as described above.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The obstacle avoidance control method of the autonomous driving vehicle in the present invention comprises the steps of: obtaining camera detection data and laser radar detection data on the autonomous driving vehicle, wherein the camera detection data includes obstacle image information, and the laser radar detection data includes obstacle distance information; processing the camera detection data by an image enhancement algorithm to obtain an obstacle area image, and processing the laser radar detection data by a point cloud clustering algorithm to obtain obstacle detection information; fusing the obstacle area image and the obstacle detection information to obtain multimodal information fusion data; and generating autonomous driving vehicle anti-collision warning information according to the multimodal information fusion data and a preset autonomous driving vehicle anti-collision warning system. According to the fused data and preset rules, the visual information and the distance information are mutually verified, which reduces misjudgment, and generates warning information to assist the autonomous driving vehicle in performing obstacle avoidance operations. The obstacle avoidance control method of the autonomous driving vehicle in the embodiment of the present invention realizes accurate detection and rapid response to obstacles through multimodal data fusion, thereby improving the safety and reliability of the autonomous driving vehicle, while also improving driving comfort and reducing the occurrence of accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of a flow chart of an obstacle avoidance control method for an autonomous driving vehicle in an embodiment of the present invention;

[0049] Figure 2 Step S in the embodiment of the present invention 300 Detailed process diagram;

[0050] Figure 3 This is a schematic diagram of an obstacle avoidance scenario for an autonomous driving vehicle in an embodiment of the present invention;

[0051] Figure 4 Schematic diagram of laser point cloud, origin coordinates and obstacle detection in an embodiment of the present invention;

[0052] Figure 5 Schematic diagram of the structure of the obstacle avoidance system of the autonomous driving vehicle in an embodiment of the present invention.

[0053] Description of reference numerals:

[0054] 100-acquisition module; 200-image processing module; 300-lidar processing module; 400-data fusion module; 500-early warning module. DETAILED DESCRIPTION

[0055] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, it can also be the internal connection of two components, it can be a wireless connection, or it can be a wired connection. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0057] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0058] See also Figure 1-5 As shown, an embodiment of the present invention provides an obstacle avoidance control method for an autonomous driving vehicle, the obstacle avoidance control method comprising the following steps:

[0059] Step S 100 : Obtain camera detection data and lidar detection data on the autonomous driving vehicle, wherein the camera detection data includes obstacle image information, and the lidar detection data includes obstacle distance information.

[0060] In this step, the system obtains real-time data from the camera and lidar, which is the basis of the obstacle avoidance system. For example, the on-board camera captures images of the surrounding environment, which contain visual information of obstacles; the lidar captures the distance information of obstacles, which provides the precise spatial position of obstacles.

[0061] Step S 200 : The camera detection data is processed by an image enhancement algorithm to obtain an obstacle area image, and the lidar detection data is processed by a point cloud clustering algorithm to obtain obstacle detection information.

[0062] In this step, the camera detection data is processed through an image enhancement algorithm to highlight obstacle features and improve recognition accuracy.

[0063] The lidar data is processed through a point cloud clustering algorithm to determine the location and distance of obstacles and extract spatial location information.

[0064] Step S 300 : Fusing the obstacle area image and the obstacle detection information to obtain multimodal information fusion data.

[0065] In this step, the image processing results (obstacle area image) and the lidar processing results (obstacle detection information) are combined to form more comprehensive obstacle information.

[0066] The system does not rely on a single sensor system. Even if one sensor system fails, the system can still detect obstacles through data from another sensor.

[0067] Step S 400 : Generate autonomous driving vehicle anti-collision warning information based on the multimodal information fusion data and a preset autonomous driving vehicle anti-collision warning system.

[0068] In this step, according to the fused data and preset rules, the visual information and distance information are mutually verified, which reduces misjudgment, and generates warning information to assist the autonomous driving vehicle in obstacle avoidance operations. The autonomous driving vehicle obstacle avoidance control method of the embodiment of the present invention realizes accurate detection and rapid response to obstacles through multimodal data fusion, thereby improving the safety and reliability of the autonomous driving vehicle, while also improving driving comfort and reducing the occurrence of accidents.

[0069] For further information, see Figure 3 As shown, in step S 200 In the method, the camera detection data is processed by an image enhancement algorithm to obtain an obstacle area image, comprising the following steps:

[0070] Acquire image data of the surrounding environment through the vehicle-mounted camera;

[0071] Preprocessing the image data, wherein the preprocessing includes denoising, enhancement, and transformation operations;

[0072] Combining the logarithmic image processing model and Lee image enhancement algorithm, the preprocessed image data is enhanced to obtain balanced image data;

[0073] Extracting features related to obstacle boundaries from the equalized image data;

[0074] Use deep learning algorithms to classify and identify the extracted features to determine the type and location of obstacles;

[0075] The obstacle image is filtered by a feature filter to obtain an obstacle area image.

[0076] Therefore, the preprocessed image data is enhanced by combining the logarithmic image processing model and the Lee image enhancement algorithm to obtain balanced image data. The logarithmic image processing model expands the dynamic range of the image and enhances the details of the dark area, while the Lee algorithm reduces noise and improves image quality and readability by using the local statistical characteristics of the image for speckle filtering. The combined use of these two algorithms can effectively improve the contrast and details of the image, making the obstacle features more obvious. Features related to the obstacle boundary are extracted from the balanced image data to improve the accuracy of obstacle recognition and provide key information for subsequent classification and recognition; the extracted features are classified and recognized using a deep learning algorithm to determine the type and location of the obstacle. The application of deep learning, especially convolutional neural network (CNN) in obstacle detection and recognition, effectively improves the recognition rate and accuracy by learning features through a large amount of training data; finally, the obstacle image is filtered by feature filters to obtain an obstacle area image, reduce background noise and interference, highlight obstacles, and improve the accuracy and robustness of obstacle detection.

[0077] It needs to be further explained here that denoising technology can reduce the noise in the image and improve the image quality; image enhancement technology can improve the contrast of the image and make the obstacle features more obvious; and image transformation (color space conversion, size adjustment, etc.) operations can be performed to meet subsequent processing requirements.

[0078] In addition, the logarithmic image processing model used in this embodiment refers to enhancing the contrast and details of an image by converting the image pixel values ​​into a logarithmic domain, which is particularly effective when processing images with poor visibility such as low light or haze. The basic principles are as follows:

[0079] Map the image pixel values ​​to the logarithmic domain, that is:

[0080]

[0081] in, is the log domain image pixel value, is the original image pixel value.

[0082] Through logarithmic transformation, the contrast of darker areas in the image is enhanced, while the contrast of brighter areas is reduced, making the details of the image clearer, thereby improving the overall visibility of the image. The logarithmic image processing model includes edge detection, image enhancement and image restoration, and is particularly suitable for enhancing the dark details of the image.

[0083] Lee image enhancement algorithm is a filtering method that is particularly suitable for denoising synthetic aperture radar (SAR) images. It performs speckle filtering based on the local statistical characteristics of the image. The following is the basic working principle of Lee image enhancement algorithm:

[0084] Lee algorithm is based on the multiplicative speckle noise model, assuming that speckle noise is uncorrelated with the image signal and that speckle noise is a stationary noise with a mean of 1. The algorithm uses the sample mean and variance within the filter window as the prior mean and variance to reduce the impact of noise.

[0085] The calculation formula of Lee filter is:

[0086]

[0087] in, represents the pixel value after filtering, Represents the pixel value of the original image, N represents the size of the filter, represents the distance to the center pixel, represents the variance of the noise, represents the sum of squares of h.

[0088] The specific implementation process of Lee's image enhancement algorithm is as follows: read the image and convert it to a grayscale image; calculate the local mean and variance, usually implemented by sliding windows; apply a filter to filter each pixel according to the mean and variance; display the original image and the filtered image for comparison

[0089] Furthermore, the calculation expression of the Lee image enhancement algorithm is:

[0090]

[0091] in: Represents the pixel brightness value of the processed image, is the pixel brightness value of the original image, So centered The arithmetic mean of the window, and is a real number, [•] + Indicates taking a positive value, that is, if the value in the brackets is negative, take 0; if it is positive, take itself.

[0092] Therefore, the Lee image enhancement algorithm effectively balances the relationship between image enhancement and noise suppression by combining local statistical characteristics and noise characteristics, thereby achieving the purpose of improving image quality.

[0093] For further information, see Figure 4 As shown, in step S 200 In the method, the laser radar detection data is processed by a point cloud clustering algorithm to obtain obstacle detection information, including:

[0094] Determine the size of the two-dimensional matrix based on the horizontal and vertical viewing angles of the laser radar, determine the coordinates of each corresponding point cloud in the two-dimensional matrix based on the distance and orientation of each laser point relative to the laser radar in the horizontal and vertical directions, and perform matrix filling on the two-dimensional matrix based on the three-dimensional coordinates of each point and its distance relative to the laser radar;

[0095] Calculate the elevation angle and elevation angle difference in the vertical direction for adjacent laser points, and mark the points where the elevation angle and elevation angle difference are both less than the corresponding threshold as ground points;

[0096] Calculate the point cloud distance for adjacent laser points;

[0097] The curvature in the horizontal direction at the location of each laser point is calculated using the point cloud polar coordinate curvature calculation formula, and the result is stored as a point cloud curvature two-dimensional matrix based on the coordinates of the point cloud in the two-dimensional matrix;

[0098] Combining the point cloud curvature two-dimensional matrix with the corresponding curvature threshold to determine whether they belong to the same obstacle category;

[0099] Based on the determined category results, breadth-first search clustering is performed on all points, and the same number is set for the same type of obstacle points and stored as a two-dimensional matrix of clustering results, thereby realizing the identification of different obstacles.

[0100] Specifically in this embodiment, the first step is to construct a two-dimensional matrix: the size of the two-dimensional matrix is ​​determined based on the horizontal and vertical viewing angles of the laser radar, which involves projecting the point cloud data in the three-dimensional space onto a two-dimensional plane for further processing.

[0101] Then, the point cloud coordinates are mapped to determine the coordinates of each point cloud in the two-dimensional matrix, and the three-dimensional coordinates of each point and its distance relative to the laser radar are filled in the two-dimensional matrix based on the coordinates, thereby converting the point cloud data into a two-dimensional matrix form.

[0102] Then, the elevation angle and elevation angle difference in the vertical direction are calculated for adjacent laser points, and the points whose elevation angle and elevation angle difference are both less than the corresponding threshold are marked as ground points, which helps to distinguish between ground and non-ground targets.

[0103] Then, the point cloud distance is calculated for adjacent laser points and the clustering features in the point cloud are identified.

[0104] Then, the curvature in the horizontal direction at the location of each laser point is calculated using the point cloud polar coordinate curvature calculation formula, and the result is stored as a point cloud curvature two-dimensional matrix to identify the edge and shape features of the obstacle.

[0105] The two-dimensional matrix of point cloud curvature and the corresponding curvature threshold are combined to determine whether they belong to the same obstacle category. The point cloud is classified based on the curvature features to identify different obstacles.

[0106] Based on the determined category results, breadth-first search clustering is performed on all points, and the same number is set for the same type of obstacle points and stored as a two-dimensional matrix of clustering results, thereby realizing the identification of different obstacles.

[0107] Therefore, through a series of preprocessing, ground point segmentation, cluster classification and breadth-first search steps, the accuracy and robustness of obstacle detection are effectively improved, providing strong technical support for the obstacle avoidance system of autonomous driving vehicles.

[0108] Furthermore, the point cloud polar coordinate curvature calculation formula is:

[0109]

[0110] in, is the horizontal angle of a point relative to the laser radar, is the elevation angle of a point relative to the laser radar, is the polar diameter based on the polar coordinate system, , Represents the polar radius of n neighboring points.

[0111] This formula can accurately calculate the curvature of each point in the point cloud. The polar coordinate curvature calculation formula is robust to noise and data missing, so it is particularly suitable for obstacles of various shapes and can handle complex environments and different obstacle types.

[0112] For further information, see Figure 4 As shown, the combining of the point cloud curvature two-dimensional matrix and the corresponding curvature threshold to determine whether they belong to the same obstacle category includes:

[0113] Calculate the deviation of the curvature of each laser point in the horizontal direction from the center of cluster curvature;

[0114] The laser radar point cloud data to be clustered whose deviation is less than or equal to the corresponding curvature threshold is used as new laser point cloud detection data to be clustered.

[0115] Specifically, the local geometric characteristics of each point are analyzed to determine the rate of change of its relative position with the surrounding points; the cluster curvature center is determined by statistical methods such as mean or median, which represents the curvature characteristics of a local area; the deviation of the curvature of each laser point from the cluster curvature center is calculated, and points with significantly different curvature from the center are identified, which may be the edges of obstacles; a curvature threshold is used to determine whether the points belong to the same obstacle category, which is used to distinguish different obstacles or ground surfaces; through curvature analysis, the outline and shape of obstacles can be more accurately identified, thereby improving the accuracy of obstacle detection.

[0116] For further information, see Figure 2 As shown, in step S 300 The step of fusing the obstacle area image and the obstacle detection information to obtain multimodal information fusion data includes:

[0117] Step S 310 : Performing data alignment on the obstacle area image and the obstacle detection information.

[0118] In this step, the purpose of data alignment is to ensure that the obstacle area image and obstacle detection information are consistent in space and time.

[0119] Step S 320 : Associating the aligned obstacle area image with the obstacle detection information, wherein the association process is used to match the same obstacles;

[0120] In this step, the data is associated by matching the same obstacles in the image and detection information.

[0121] Step S 330 : Generate environmental perception information as the multimodal information fusion data based on the association result, the three-dimensional object position information in the obstacle area image, and the obstacle distance, obstacle movement speed and obstacle movement angle of the obstacle detection information.

[0122] In this step, the environmental perception information is generated by combining the association results and the information in the two data sources (such as three-dimensional position, distance, speed, and angle).

[0123] In summary, the technical principles and processes of point cloud curvature processing and multimodal information fusion are aimed at improving the accuracy and robustness of obstacle detection. They provide more comprehensive environmental information by combining data from different sensors, thereby enhancing the environmental perception capability of the autonomous driving system.

[0124] See also Figure 5 As shown, an embodiment of the present invention further provides an obstacle avoidance system for an autonomous driving vehicle, the obstacle avoidance system for an autonomous driving vehicle comprising:

[0125] An acquisition module 100 is used to acquire camera detection data and laser radar detection data, wherein the camera detection data includes obstacle image information and the laser radar detection data includes obstacle distance information;

[0126] The image processing module 200 is used to process the camera detection data through an image enhancement algorithm to obtain an obstacle area image;

[0127] The laser radar processing module 300 is used to process the laser radar detection data by using a point cloud clustering algorithm to obtain obstacle detection information;

[0128] A data fusion module 400 is used to fuse the obstacle area image and the obstacle detection information to obtain multi-modal information fusion data;

[0129] The warning module 500 is used to generate autonomous driving vehicle anti-collision warning information based on the multimodal information fusion data and a preset autonomous driving vehicle anti-collision warning system.

[0130] Specifically, the autonomous driving vehicle obstacle avoidance system described in the embodiment of the present invention is a multimodal fusion system that comprehensively utilizes data from cameras and LiDAR to improve the accuracy and robustness of obstacle detection.

[0131] The acquisition module 100 is used to acquire an environment image, including visual information of obstacles and distance information of obstacles.

[0132] The image processing module 200 is used to process the camera detection data and enhance the features of obstacles in the image to make them easier to identify.

[0133] The laser radar processing module 300 is used to process the laser radar data, identify obstacles and vehicles, etc. through clustering algorithm analysis, and extract the spatial distance information of the obstacles.

[0134] The data fusion module 400 is used to combine the results of the image processing module 200 and the laser radar processing module 300 to form more comprehensive obstacle information and improve the accuracy and robustness of the system.

[0135] The warning module 500 is used to generate warning information based on the fused data and warning rules to assist the autonomous driving vehicle in performing obstacle avoidance operations.

[0136] It should be noted that it should be understood that the division of the various modules of the above obstacle avoidance system is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or physically separated. Moreover, these modules can be all implemented in the form of software calling through processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software calling through processing elements, and some modules can be implemented in the form of hardware.

[0137] For example, the image processing module 200 may be a separate processing element, or may be integrated into a chip of the obstacle avoidance system, or may be stored in the memory of the obstacle avoidance system in the form of program code, and may be called and executed by a processing element of the obstacle avoidance system. The implementation of other modules is similar.

[0138] In addition, all or part of these modules can be integrated together or implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed by an integrated logic circuit of hardware in the processor element or an instruction in the form of software.

[0139] Another embodiment of the present invention also provides an autonomous driving vehicle, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the autonomous driving vehicle obstacle avoidance control method described above.

[0140] The autonomous driving car in this embodiment includes: a processor and a memory, wherein the processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor may be implemented in at least one hardware form of DSP, FPGA, and PLA.

[0141] The processor may also include a main processor and a coprocessor. The main processor is a processor used to process data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor used to process data in the standby state.

[0142] In some embodiments, the processor may be integrated with a GPU, and the GPU is responsible for rendering and drawing the content that needs to be displayed on the display screen.

[0143] In other embodiments, the processor may also include an AI processor for processing computing operations related to machine learning.

[0144] The memory may include one or more computer-readable storage media, which may be non-transitory. The memory may also include high-speed random access memory, and non-volatile memory, such as one or more disk storage devices, flash memory storage devices.

[0145] In some embodiments, the non-transitory computer-readable storage medium in the memory is used to store at least one instruction, which is used to be executed by the processor to implement the autonomous driving vehicle obstacle avoidance control method provided by the method embodiment of the present application.

[0146] To achieve the above objectives and other related objectives, a computer-readable storage medium may also be provided in an embodiment of the present invention, including: a memory and a processor; the memory stores a computer program, and the processor runs the computer program to execute some or all of the steps in the aforementioned method embodiment.

[0147] Specifically, the memory may include a random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as a ROM or a disk memory.

[0148] The processor may include a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0149] A person skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to the computer program. The aforementioned computer program can be stored in a computer-readable storage medium. When the program is executed, it executes some or all of the steps of the automatic driving vehicle obstacle avoidance control method in the above-mentioned embodiment; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0150] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. An obstacle avoidance control method for an autonomous driving vehicle, characterized in that: The following steps are involved: Step S 100 : Obtaining camera detection data and laser radar detection data on the autonomous driving vehicle, wherein the camera detection data includes obstacle image information, and the laser radar detection data includes obstacle distance information; Step S 200 : Processing the camera detection data through an image enhancement algorithm to obtain an obstacle area image, and processing the laser radar detection data through a point cloud clustering algorithm to obtain obstacle detection information; The step of processing the laser radar detection data by using a point cloud clustering algorithm to obtain obstacle detection information includes: Determine the size of the two-dimensional matrix based on the horizontal and vertical viewing angles of the laser radar, determine the coordinates of each corresponding point cloud in the two-dimensional matrix based on the distance and orientation of each laser point relative to the laser radar in the horizontal and vertical directions, and perform matrix filling on the two-dimensional matrix based on the three-dimensional coordinates of each point and its distance relative to the laser radar; Calculate the elevation angle and elevation angle difference in the vertical direction for adjacent laser points, and mark the points where the elevation angle and elevation angle difference are both less than the corresponding threshold as ground points; Calculate the point cloud distance for adjacent laser points; The curvature in the horizontal direction at the location of each laser point is calculated using the point cloud polar coordinate curvature calculation formula, and the result is stored as a point cloud curvature two-dimensional matrix based on the coordinates of the point cloud in the two-dimensional matrix; Combining the point cloud curvature two-dimensional matrix with the corresponding curvature threshold to determine whether they belong to the same obstacle category; Based on the determined category results, a breadth-first search clustering is performed on all points, and the same number is set for the same type of obstacle points and stored as a two-dimensional matrix of clustering results, thereby realizing the identification of different obstacles; Step S 300 : Fusing the obstacle area image and the obstacle detection information to obtain multimodal information fusion data; The step of fusing the obstacle area image and the obstacle detection information to obtain multimodal information fusion data includes: Step S 310 : Performing data alignment on the obstacle area image and the obstacle detection information; Step S 320 : Associating the aligned obstacle area image with the obstacle detection information, wherein the association process is used to match the same obstacles; Step S 330 : Generate environmental perception information as the multimodal information fusion data according to the association result, the three-dimensional object position information in the obstacle area image, and the obstacle distance, obstacle moving speed and obstacle moving angle of the obstacle detection information; Step S 400 : Generate autonomous driving vehicle anti-collision warning information based on the multimodal information fusion data and a preset autonomous driving vehicle anti-collision warning system.

2. The obstacle avoidance control method for an autonomous driving vehicle according to claim 1, characterized in that: In step S 200 In the method, the camera detection data is processed by an image enhancement algorithm to obtain an obstacle area image, comprising the following steps: Acquire image data of the surrounding environment through the vehicle-mounted camera; Preprocessing the image data, wherein the preprocessing includes denoising, enhancement, and transformation operations; Combining the logarithmic image processing model and Lee image enhancement algorithm, the preprocessed image data is enhanced to obtain balanced image data; Extracting features related to obstacle boundaries from the equalized image data; Use deep learning algorithms to classify and identify the extracted features to determine the type and location of obstacles; The obstacle image is filtered by a feature filter to obtain an obstacle area image.

3. The obstacle avoidance control method for an autonomous driving vehicle according to claim 2, characterized in that: The calculation expression of the Lee image enhancement algorithm is: in: Represents the pixel brightness value of the processed image, is the pixel brightness value of the original image, So centered The arithmetic mean of the window, and is a real number, [•] + Indicates taking a positive value, that is, if the value in the brackets is negative, take 0; if it is positive, take itself.

4. The obstacle avoidance control method for an autonomous driving vehicle according to claim 1, characterized in that: The point cloud polar coordinate curvature calculation formula is: in, is the horizontal angle of a point relative to the laser radar, is the elevation angle of a point relative to the laser radar, is the polar diameter based on the polar coordinate system, , Represents the polar radius of n neighboring points.

5. The obstacle avoidance control method for an autonomous driving vehicle according to claim 4, characterized in that: The combining the point cloud curvature two-dimensional matrix with the corresponding curvature threshold to determine whether the objects belong to the same obstacle category includes: Calculate the deviation of the curvature of each laser point in the horizontal direction from the center of cluster curvature; The laser radar point cloud data to be clustered whose deviation is less than or equal to the corresponding curvature threshold is used as new laser point cloud detection data to be clustered.

6. An obstacle avoidance system for an autonomous vehicle, characterized in that: include: An acquisition module, used to acquire camera detection data and laser radar detection data on the autonomous driving vehicle, wherein the camera detection data includes obstacle image information, and the laser radar detection data includes obstacle distance information; An image processing module is used to process the camera detection data through an image enhancement algorithm to obtain an obstacle area image; A laser radar processing module, used to process the laser radar detection data through a point cloud clustering algorithm to obtain obstacle detection information; The step of processing the laser radar detection data by a point cloud clustering algorithm to obtain obstacle detection information specifically includes: Determine the size of the two-dimensional matrix based on the horizontal and vertical viewing angles of the laser radar, determine the coordinates of each corresponding point cloud in the two-dimensional matrix based on the distance and orientation of each laser point relative to the laser radar in the horizontal and vertical directions, and perform matrix filling on the two-dimensional matrix based on the three-dimensional coordinates of each point and its distance relative to the laser radar; Calculate the elevation angle and elevation angle difference in the vertical direction for adjacent laser points, and mark the points where the elevation angle and elevation angle difference are both less than the corresponding threshold as ground points; Calculate the point cloud distance for adjacent laser points; The curvature in the horizontal direction at the location of each laser point is calculated using the point cloud polar coordinate curvature calculation formula, and the result is stored as a point cloud curvature two-dimensional matrix based on the coordinates of the point cloud in the two-dimensional matrix; Combining the point cloud curvature two-dimensional matrix with the corresponding curvature threshold to determine whether they belong to the same obstacle category; Based on the determined category results, a breadth-first search clustering is performed on all points, and the same number is set for the same type of obstacle points and stored as a two-dimensional matrix of clustering results, thereby realizing the identification of different obstacles; A data fusion module, used for fusing the obstacle area image and the obstacle detection information to obtain multi-modal information fusion data; The step of fusing the obstacle area image and the obstacle detection information to obtain multimodal information fusion data specifically includes: Aligning the obstacle area image and the obstacle detection information; associating the aligned obstacle area image and the obstacle detection information, wherein the association process is used to match the same obstacles; generating environmental perception information as the multimodal information fusion data according to the association result, the three-dimensional object position information in the obstacle area image, and the obstacle distance, obstacle moving speed, and obstacle moving angle of the obstacle detection information; The warning module is used to generate the anti-collision warning information of the autonomous driving vehicle according to the multimodal information fusion data and the preset anti-collision warning system of the autonomous driving vehicle.

7. An automatic driving car, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the obstacle avoidance control method for an autonomous driving vehicle as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the obstacle avoidance control method for an autonomous driving vehicle as described in any one of claims 1 to 5.

Images