Dynamic Point Cloud Detection Method and System Based on Elevation Image

Abstract

The present invention belongs to the technical field of lidar. A dynamic point cloud detection method and system based on elevation images are proposed, which obtain the ground point cloud, non-ground point cloud and odometer information of each frame of lidar point cloud; use the static point cloud and ground point cloud obtained from the previous frame to construct an elevation image; perform ground continuity repair on the elevation image; combine the odometer information, and perform occlusion detection on the point cloud of the current frame based on the elevation image to determine dynamic points, suspected dynamic points and suspected static points; perform regional consistency check on the pixels where the dynamic points, the suspected dynamic points and the suspected static points are located to determine the final dynamic points; perform clustering operation on the detected final dynamic point cloud to obtain the dynamic point cloud detection result. The present invention effectively overcomes problems such as the lack of map background point cloud, holes in depth images, and the easy misjudgment of ground point cloud, thereby significantly improving the accuracy of dynamic object detection.

Description

Technical Field

[0001] The present invention relates to the technical field of lidar, and particularly to a dynamic point cloud detection method and system based on elevation images. Background Art

[0002] The statements in this part merely provide background art related to the present invention and do not necessarily constitute prior art.

[0003] With the development of unmanned technology, autonomous robots have shown transformative potential in multiple fields, such as driverless taxis, agricultural robots, and factory handling robots, making them increasingly important in daily life. However, when deploying these robots in the actual environment, detecting and avoiding unknown moving objects pose a challenge. For example, a suddenly crossing pedestrian or cyclist may collide with an autonomous vehicle.

[0004] To avoid such accidents, autonomous robots usually rely on the lidar carried to detect moving objects. Currently, lidar-based dynamic object detection methods have become a research hotspot, and important progress has been made in related research, mainly divided into three categories: point cloud segmentation methods, ray tracing methods, and visibility methods.

[0005] Traditional point cloud segmentation methods are mainly achieved through clustering. This type of method has good dynamic object detection effects, but its computational cost is relatively high. Therefore, semantic segmentation methods based on deep learning have gradually become a popular research object. Although this method is simple and efficient, its model performance highly depends on the quality of the pre-trained dataset, with certain limitations.

[0006] Ray tracing methods are based on the principle of ray propagation. By traversing voxels, the end points of rays are marked as occupied voxels, and the space through which the rays pass is marked as free voxels. Although this method has good theoretical support, it requires constructing a huge voxel map, consuming a large amount of computing and memory resources. In addition, this method depends on extremely accurate positioning information, and its application scope is limited.

[0007] Visibility methods are based on the principle of viewpoint occlusion. By constructing a depth image of map points, it is determined whether the current point occludes the map, thereby realizing the detection of dynamic objects. This type of method has the advantages of small computational amount and high speed, and at the same time does not require constructing a large-scale voxel map, so it has a wide application scope.

[0008] Although visibility methods perform excellently in dynamic object detection, there are certain limitations to this type of method: when there is a lack of background map point clouds, it is difficult for this type of method to effectively detect dynamic objects, which limits its applicability in some scenarios; due to the lack of continuity of background objects, this type of method generally uses linear interpolation methods to fill the holes in depth images, which also leads to errors; in addition, ground point clouds usually have a large incident angle, which makes this type of method prone to misjudging ground point clouds as dynamic objects. Summary of the Invention

[0009] To address the deficiencies of the prior art, the present invention provides a method and system for dynamic point cloud detection based on elevation images, effectively overcoming problems such as the lack of map background point clouds, holes in depth images, and the easy misjudgment of ground point clouds, thereby significantly improving the accuracy of dynamic object detection.

[0010] To achieve the above objective, the present invention adopts the following technical solutions:

[0011] In a first aspect, the present invention provides a method for dynamic point cloud detection based on elevation images.

[0012] A method for dynamic point cloud detection based on elevation images includes the following processes:

[0013] Obtain the ground point cloud, non-ground point cloud, and odometer information of each frame of lidar point cloud;

[0014] Construct an elevation image using the static point cloud and ground point cloud obtained from the previous frame;

[0015] Perform ground continuity repair on the elevation image;

[0016] Combined with the odometer information, perform occlusion detection on the point cloud of the current frame based on the elevation image to determine dynamic points, suspected dynamic points, and suspected static points;

[0017] Perform regional consistency checks on the pixels where the dynamic points, suspected dynamic points, and suspected static points are located to determine the final dynamic points;

[0018] Perform clustering operations on the detected final dynamic point cloud to obtain the dynamic point cloud detection result.

[0019] In a second aspect, the present invention provides a system for dynamic point cloud detection based on elevation images.

[0020] A system for dynamic point cloud detection based on elevation images includes:

[0021] A data acquisition unit configured to: obtain the ground point cloud, non-ground point cloud, and odometer information of each frame of lidar point cloud;

[0022] An elevation image construction unit, configured to: construct an elevation image by using the static point cloud and the ground point cloud obtained from the previous frame;

[0023] A ground continuity repair unit, configured to: repair the ground continuity of the elevation image;

[0024] An occlusion detection unit, configured to: combine the odometer information, and perform occlusion detection on the point cloud of the current frame based on the elevation image to determine dynamic points, suspected dynamic points, and suspected static points;

[0025] A regional consistency check unit, configured to: perform a regional consistency check on the pixels where the dynamic points, the suspected dynamic points, and the suspected static points are located to determine the final dynamic points;

[0026] A detection result generation unit, configured to: perform a clustering operation on the detected final dynamic point cloud to obtain a dynamic point cloud detection result.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] 1. The present invention adopts an elevation image method based on a concentric circle model. By changing the projection direction of the traditional depth image in the visibility method, problems such as the lack of map background point cloud and the misjudgment of ground point cloud are avoided, thereby significantly improving the accuracy of dynamic object detection.

[0029] 2. The present invention proposes a method for ground continuity repair based on the assumption of ground point cloud continuity, effectively fills the hole pixels in the elevation image, increases the ground pixels available for occlusion detection, effectively reduces the misjudgment rate and missed detection rate of dynamic point cloud, and improves the reliability of the detection result.

[0030] 3. The present invention introduces a multi-frame occlusion detection and regional consistency check mechanism, effectively reduces the requirements for computing and memory resources, further improves the robustness and real-time performance of dynamic detection, and is applicable to resource-constrained embedded devices and applications in complex dynamic environments.

[0031] The advantages of the additional aspects of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The specification drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0033] Figure 1 It is a schematic flowchart of a dynamic point cloud detection method based on an elevation image provided in Embodiment 1 of the present invention;

[0034] Figure 2 Schematic diagram of the data acquisition method provided in Embodiment 1 of the present invention;

[0035] Figure 3 Schematic diagram of the elevation image provided in Embodiment 1 of the present invention;

[0036] Figure 4 Schematic diagram of the ground continuity repair method provided in Embodiment 1 of the present invention;

[0037] Figure 5 Schematic diagram of the occlusion detection principle provided in Embodiment 1 of the present invention;

[0038] Figure 6 Schematic diagram of the regional consistency check principle provided in Embodiment 1 of the present invention;

[0039] Figure 7 Schematic diagram of a dynamic point cloud detection system based on an elevation image provided in Embodiment 2 of the present invention. Detailed implementation manners

[0040] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0041] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0042] In the case of no conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0043] Embodiment 1:

[0044] This implementation manner proposes a dynamic point cloud detection method based on an elevation image, as Figure 1 shown, including the following processes:

[0045] S1: Obtain the ground point cloud, non-ground point cloud and odometer information of each frame of lidar point cloud;

[0046] S2: Construct an elevation image using the static point cloud and ground point cloud obtained in the previous frame;

[0047] S3: Perform ground continuity repair on the elevation image;

[0048] S4: Combine the odometer information, and perform occlusion detection on the point cloud of the current frame based on the elevation image to determine dynamic points, suspected dynamic points and suspected static points;

[0049] S5: Perform a regional consistency check on the pixels where the dynamic points, the suspected dynamic points, and the suspected static points are located to determine the final dynamic points;

[0050] S6: Perform a clustering operation on the detected final dynamic point cloud to obtain the dynamic point cloud detection result.

[0051] In S1 of this implementation method, specifically, it includes:

[0052] Receive the ground point cloud and non - ground point cloud output by the ground point cloud segmentation system, and at the same time obtain the odometry data provided by the multi - sensor fusion simultaneous localization and mapping system to complete the acquisition of input data. As Figure 2 shown, the obtained odometry data refers to the relative pose between two frames of lidar point clouds. The multi - sensor fusion simultaneous localization and mapping system receives lidar point clouds and IMU data, outputs odometry data and a point cloud map, outputs the undistorted point cloud to the ground point cloud segmentation system, and outputs the odometry data to the dynamic point cloud detection system based on the elevation image. The ground point cloud segmentation system conveys the ground point cloud and non - ground point cloud to the dynamic point cloud detection system based on the elevation image. After generating the ground point cloud, static point cloud, and dynamic point cloud, the dynamic point cloud detection system based on the elevation image outputs them to the multi - sensor fusion simultaneous localization and mapping system.

[0053] In S2 of this implementation method, specifically, it includes:

[0054] The elevation image consists of four circular regions, namely circular region one, circular region two, circular region three, and circular region four. As Figure 3 shown, the pixel density of these circular regions gradually decreases from the inside to the outside, forming a distribution pattern from dense to sparse. The radial radius formulas for the four circular regions are:

[0055] (1);

[0056] Among them, the outer diameter of circular region one is , and the inner diameter is ; the outer diameter of circular region two is , and the inner diameter is ; the outer diameter of circular region three is , and the inner diameter is ; the outer diameter of circular region four is , and the inner diameter is , The meaning of

[0057] The pixel resolutions of the four circular regions are respectively:

[0058] (2);

[0059] Among them, the radial resolution of the first annular region is , and the angular resolution is . The number of radial rings in this region is , , and the number of sectors in each ring is , ;

[0060] The radial resolution of the second annular region is , and the angular resolution is . The number of radial rings in this region is , , and the number of sectors in each ring is , ;

[0061] The radial resolution of the third annular region is , and the angular resolution is . The number of radial rings in this region is , , and the number of sectors in each ring is , ;

[0062] The radial resolution of the fourth annular region is , and the angular resolution is . The number of radial rings in this region is , , and the number of sectors in each ring is , .

[0063] The elevation image is constructed based on the static point cloud and the ground point cloud obtained from the previous frame. First, each point in the static point cloud and the ground point cloud is converted from the Cartesian coordinate system to the polar coordinate system , and the conversion formula is as follows:

[0064] (3);

[0065] (4).

[0066] Then, according to the resolution of the elevation image, each point is mapped to the corresponding elevation image pixel, and the index of each pixel is composed of the ring region number and the sector number . The formula for calculating the pixel index where each point is located is as follows:

[0067] (5);

[0068] In the elevation image, each pixel records the following information: the number of point clouds contained in the pixel, the coordinate values of the point clouds, and the type of the point clouds. If the pixel contains ground point clouds, the pixel is marked as a ground pixel; otherwise, it is a non-ground pixel.

[0069] In step S3 of this implementation manner, based on the assumption of ground point cloud continuity, ground continuity repair is performed on non-ground pixels in the elevation image. Specifically, it includes:

[0070] S3.1: The repair process starts from circular region 1. First, find all ground pixels in this region. Traverse in descending order of the circular region numbers of the pixels. Starting from each ground pixel, search for new ground pixels pixel by pixel along the ray direction towards the center of the elevation image;

[0071] If new ground pixels are found in circular region 1 or circular region 2, perform ground continuity repair on all non-ground pixels between the starting point and the new ground pixels, and mark them as ground-repaired pixels;

[0072] If no new ground pixels are found in circular region 2, stop traversing this ray and do not perform ground continuity repair on the traversed pixels.

[0073] S3.2: After the traversal of ground pixels in circular region 1 is completed, continue to process circular region 2. Extract all ground pixels in this region, traverse in descending order of the circular region numbers of the pixels. Starting from each ground pixel, search for new ground pixels pixel by pixel along the ray direction towards the center of the elevation image;

[0074] If new ground pixels are found in circular region 2 or circular region 3, perform ground continuity repair on all non-ground pixels between the starting point and the new ground pixels, and mark them as ground-repaired pixels;

[0075] If no new ground pixels are found in circular region 3, stop traversing this ray and do not perform ground continuity repair on the traversed pixels.

[0076] S3.3: After the traversal of ground pixels in circular region 2 is completed, continue to process circular region 3. Extract all ground pixels in this region, traverse in descending order of the circular region numbers of the pixels. Starting from each ground pixel, search for new ground pixels pixel by pixel along the ray direction towards the center of the elevation image;

[0077] If new ground pixels are found in circular region 3 or circular region 4, perform ground continuity repair on all non-ground pixels between the starting point and the new ground pixels, and mark them as ground-repaired pixels;

[0078] If no new ground pixels are found in circular region 4, stop traversing this ray and do not perform ground continuity repair on the traversed pixels.

[0079] S3.4: After the traversal of the ground pixels in the third annular region is completed, continue to process the fourth annular region. Extract all the ground pixels within this region, traverse them in descending order according to the pixel annular region number, and starting from each ground pixel, along the ray direction towards the center of the elevation image, perform ground continuity repair on all non-ground pixels passed by the rays, and mark these pixels as ground repair pixels.

[0080] At this point, the ground continuity repair of the elevation image is completed, and all the possible missing ground point clouds in the non-ground pixels have been supplemented, as Figure 4 shown.

[0081] Ground continuity repair method: According to the distance between the ground repair pixel and the starting ground pixel and the found ground pixel, use the linear interpolation method to calculate the ground point cloud height of the ground repair pixel, and its calculation formula is as follows:

[0082] (6);

[0083] Among them, the ground height of the pixel to be repaired is , the ground height of the starting ground pixel is , the ground height of the found ground pixel is , the distance from the pixel to be repaired to the starting ground pixel is , and the total distance between the starting ground pixel and the found ground pixel is .

[0084] In S4 of this implementation method, specifically, it includes:

[0085] Always maintain the elevation image of the last five frames of lidar point clouds. Therefore, in the system initialization stage, a time window of the first five frames of lidar point clouds is required. During this period, all non-ground point clouds are regarded as static point clouds, and the point clouds are only used to construct the elevation image without occlusion detection.

[0086] Starting from the sixth frame, perform occlusion detection on the point clouds. The specific method is as follows: First, use the odometry data to transform each non-ground point cloud of the current frame into the coordinate system of the elevation image of the first five frames one by one, and the transformation formula is as follows:

[0087] (7);

[0088] Among them, the current frame lidar point cloud is , the point cloud of the current frame lidar point cloud transformed into the coordinate system of the elevation image of the first five frames is , and the odometry data of the first five frames is .

[0089] Then, the transformed point cloud is divided into corresponding pixels according to the resolution of the elevation image. The division method is as follows: first, convert the points from the Cartesian coordinate system to the polar coordinate system, and then map each point to the corresponding pixel according to the resolution of the elevation image.

[0090] Next, perform occlusion detection on the pixels where the points mapped to the elevation image are located. The principle of occlusion detection is as Figure 5 shown. If the difference between the Z-axis height of the point to be detected and the Z-axis height of the highest ground point in the pixel exceeds the set threshold , it is considered that the point to be detected occludes the ground point; otherwise, it is considered that no occlusion occurs. Similarly, if the difference between the Z-axis height of the static point in the pixel and the Z-axis height of the point to be detected exceeds the threshold , it is considered that the point to be detected is occluded by the static point; otherwise, it is considered that no occlusion occurs.

[0091] During the occlusion detection process, the following situations may occur:

[0092] (1) The pixel is a non-ground pixel and there is no point cloud in the pixel: The mapped point is determined to be a suspected static point;

[0093] (2) The pixel is a non-ground pixel and the mapped point occludes the static object in the pixel: The mapped point is determined to be a static point;

[0094] (3) The pixel is a non-ground pixel and the mapped point is occluded by the static object in the pixel: The mapped point is determined to be a static point;

[0095] (4) The pixel is a ground pixel or a ground repair pixel and the mapped point is occluded by the ground point in the pixel: The mapped point is determined to be an invalid point;

[0096] (5) The pixel is a ground pixel or a ground repair pixel and the mapped point only occludes the ground point: The mapped point is determined to be a dynamic point;

[0097] (6) The pixel is a ground pixel or a ground repair pixel and the mapped point both occludes the ground point and is occluded by the static point: The mapped point is determined to be a suspected dynamic point;

[0098] After completing the occlusion detection of the current frame elevation image, all mapped points will be determined to be one of the following five types: static point, dynamic point, suspected dynamic point, suspected static point, or invalid point;

[0099] Finally, integrate the occlusion detection results of the elevation images of the previous five frames. If the number of times a certain mapped point is determined to be the same type in five frames exceeds three times, it is finally marked as that type; otherwise, it is marked as a static point.

[0100] In step S5 of this implementation method, specifically, it includes:

[0101] The regional consistency check is based on the principle that the dynamic point cloud should be far from the static map point cloud, and is used to recover misjudged point clouds. This process performs a regional consistency check on the pixels marked as dynamic points, suspected dynamic points, and suspected static points that are occluded and detected in the elevation images of the first five frames. When checking, the neighborhood of the current pixel is defined as the 8-pixel area around it. The method of the regional consistency check is to determine whether there are points that meet the conditions among the neighborhood pixels. If the Z-axis height difference between a certain type of point in the neighborhood pixels and the marked point is less than the set threshold , then the type of the marked point is modified according to specific rules, and its principle is as Figure 6 shown, including:

[0102] S5.1: When the marked point is a dynamic point, if there are static points that meet the conditions among the neighborhood pixels, and the number of times this condition is met in the five-frame elevation images exceeds three times, then the marked point is re-determined as a static point; otherwise, the marked point is still determined as a dynamic point;

[0103] S5.2: When the marked point is a suspected dynamic point, if there are static points that meet the conditions among the neighborhood pixels, and the number of times this condition is met in the five-frame elevation images exceeds two times, then the marked point is re-determined as a static point; otherwise, the marked point is determined as a dynamic point;

[0104] S5.3: When the marked point is a suspected static point, if there are static points that meet the conditions among the neighborhood pixels, and the number of times this condition is met in the five-frame elevation images exceeds one time, then the marked point is re-determined as a static point; otherwise, the marked point is determined as a dynamic point.

[0105] In S6 of this implementation manner, specifically, it includes:

[0106] Perform a clustering operation on the point cloud determined to be a dynamic point after passing the regional consistency check. Use the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. After clustering is completed, the result is output as the dynamic point cloud.

[0107] Embodiment 2:

[0108] As Figure 7 shown, this implementation manner provides a dynamic point cloud detection system based on elevation images, including:

[0109] A data acquisition unit, configured to: acquire the ground point cloud, non-ground point cloud, and odometer information of each frame of lidar point cloud. The specific working method is shown in S1 in Embodiment 1:

[0110] An elevation image construction unit, configured to: construct an elevation image using the static point cloud and ground point cloud obtained from the previous frame. The specific working method is shown in S2 in Embodiment 1:

[0111] The ground continuity repair unit is configured to repair the ground continuity of the elevation image. For the specific working method, see S3 in Embodiment 1:

[0112] The occlusion detection unit is configured to, in combination with the odometer information, perform occlusion detection on the point cloud of the current frame based on the elevation image to determine dynamic points, suspected dynamic points, and suspected static points. For the specific working method, see S4 in Embodiment 1:

[0113] The regional consistency check unit is configured to perform a regional consistency check on the pixels where the dynamic points, the suspected dynamic points, and the suspected static points are located to determine the final dynamic points. For the specific working method, see S5 in Embodiment 1:

[0114] The detection result generation unit is configured to perform a clustering operation on the detected final dynamic point cloud to obtain a dynamic point cloud detection result. For the specific working method, see S6 in Embodiment 1.

[0115] It can be understood that the above-mentioned various units can be respectively or all combined into one or several other units to form, or some of them can be further split into multiple smaller units with functional division to form, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above units are divided based on logical functions. In practical applications, the function of one unit can also be realized by multiple units, or the functions of multiple units can be realized by one unit. In other embodiments of the present application, the system may also include other units. In practical applications, these functions can also be assisted by other units and can be realized by the cooperation of multiple units.

[0116] According to another embodiment of the present application, the system described in this embodiment can be constructed by running a computer program (including program code) that can execute the respective steps involved in the corresponding method described in Embodiment 1 on a general computing device such as a computer including processing elements and storage elements such as a central processing unit (CPU), a random access memory (RAM), and a read-only memory (ROM), and the method of Embodiment 1 of the present application can be realized. The computer program can be recorded on a computer-readable recording medium, loaded into the above-mentioned computing device through the computer-readable recording medium, and run therein.

[0117] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this application can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0118] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions according to the embodiments of this application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired manner (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that the computer can access or a data processing device such as a server or data center that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0119] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, 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 dynamic point cloud detection method based on elevation images, characterized in that, It includes the following processes: Obtain the ground point cloud, non-ground point cloud, and odometry information of each frame of lidar point cloud; Construct an elevation image using the static point cloud and ground point cloud obtained from the previous frame, including: converting each point in the static point cloud and the ground point cloud from the Cartesian coordinate system to the polar coordinate system, and mapping each point to the corresponding elevation image pixel according to the resolution of the elevation image. The index of each pixel consists of a ring area number and a sector number. The elevation image includes four annular regions, namely annular region one, annular region two, annular region three, and annular region four, and the pixel density gradually decreases from the inside to the outside; perform ground continuity repair on the elevation image; Combined with the odometry information, perform occlusion detection on the point cloud of the current frame based on the elevation image to determine dynamic points, suspected dynamic points, and suspected static points; Perform regional consistency check on the pixels where the dynamic points, the suspected dynamic points, and the suspected static points are located to determine the final dynamic points; Perform a clustering operation on the detected final dynamic point cloud to obtain the dynamic point cloud detection result; Among them, combined with the odometry information, perform occlusion detection on the point cloud of the current frame based on the elevation image to determine dynamic points, suspected dynamic points, and suspected static points, including: Always maintain the elevation images of the nearest M frames of lidar point cloud. In the initialization stage, a time window of the first five frames of lidar point cloud is required. During this period, all non-ground point clouds are regarded as static point clouds, and the point clouds are only used to construct the elevation image without occlusion detection. Starting from the (M + 1)-th frame, occlusion detection is performed on the point cloud; Use the odometry data to transform the non-ground point cloud of the current frame one by one into the coordinate system of the elevation images of the first five frames; Divide the transformed point cloud into corresponding pixels according to the resolution of the corresponding elevation image, including: first convert the point from the Cartesian coordinate system to the polar coordinate system, and then map each point to the corresponding pixel according to the resolution of the elevation image; Perform occlusion detection on the pixel where the point to be detected mapped in the elevation image. If the difference between the Z-axis height of the point to be detected and the Z-axis height of the highest ground point in the pixel exceeds the set threshold , it is considered that the point to be detected occludes the ground point; otherwise, it is considered that no occlusion occurs. If the difference between the Z-axis height of the static point in the pixel and the Z-axis height of the point to be detected exceeds the threshold , it is considered that the point to be detected is occluded by the static point; otherwise, it is considered that no occlusion occurs.

2. The method for detecting dynamic point cloud based on elevation image according to claim 1, wherein In the elevation image, each pixel records the following information: the number of point clouds contained in the pixel, the coordinate values of the point clouds, and the types of the point clouds; if the pixel contains ground point cloud, the pixel is marked as a ground pixel; otherwise, it is a non-ground pixel.

3. The method for detecting dynamic point cloud based on elevation image according to claim 1, wherein Performing ground continuity repair on the elevation image includes: The repair process starts from annular region one, finds all ground pixels in annular region one, traverses them in descending order of the ring area number of the pixels, and takes each ground pixel as the starting point to search for new ground pixels pixel by pixel along the ray direction towards the center of the elevation image; If new ground pixels are found in annular region one or annular region two, perform ground continuity repair on all non-ground pixels between the starting point and the new ground pixels, and mark them as ground repair pixels; If no new ground pixels are found in annular region two, stop the ray traversal and do not perform ground continuity repair on the traversed pixels; After the traversal of the ground pixels in the first annular region is completed, continue to process the second annular region. Extract all the ground pixels within the second annular region and traverse them in descending order of the pixel annular region number. Starting from each ground pixel, search for new ground pixels pixel by pixel along the ray direction towards the center of the elevation image; If new ground pixels are found in the second or third annular region, perform ground continuity repair on all non-ground pixels between the starting point and the new ground pixels, and mark them as ground repair pixels; If no new ground pixels are found in the third annular region, stop the ray traversal and do not perform ground continuity repair on the traversed pixels; After the traversal of the ground pixels in the second annular region is completed, continue to process the third annular region. Extract all the ground pixels within the third annular region and traverse them in descending order of the pixel annular region number. Starting from each ground pixel, search for new ground pixels pixel by pixel along the ray direction towards the center of the elevation image; If new ground pixels are found in the third or fourth annular region, perform ground continuity repair on all non-ground pixels between the starting point and the new ground pixels, and mark them as ground repair pixels; If no new ground pixels are found in the fourth annular region, stop the ray traversal and do not perform ground continuity repair on the traversed pixels; After the traversal of the ground pixels in the third annular region is completed, continue to process the fourth annular region. Extract all the ground pixels within the fourth annular region and traverse them in descending order of the pixel annular region number. Starting from each ground pixel, along the ray direction towards the center of the elevation image, perform ground continuity repair on all non-ground pixels passed by the rays, and mark these pixels as ground repair pixels.

4. The dynamic point cloud detection method based on elevation image according to claim 3, characterized in that Performing ground continuity repair includes: According to the distance between the ground-repaired pixel and the starting ground pixel and the found ground pixel, the ground point cloud height of the ground-repaired pixel is calculated by using the linear interpolation method, and its calculation formula is as follows: , where the ground height of the pixel to be repaired is , the ground height of the starting ground pixel is , the ground height of the found ground pixel is , the distance from the pixel to be repaired to the starting ground pixel is , the total distance between the starting ground pixel and the found ground pixel is .

5. The dynamic point cloud detection method based on elevation image according to claim 1, characterized in that During the occlusion detection process, the following situations occur: The pixel is a non-ground pixel and there is no point cloud in the pixel: the mapped point is determined to be a suspected static point; The pixel is a non-ground pixel and the mapped point occludes a static object in the pixel: the mapped point is determined to be a static point; The pixel is a non-ground pixel and the mapped point is occluded by a static object in the pixel: the mapped point is determined to be a static point; The pixel is a ground pixel or a ground repair pixel and the mapped point is occluded by a ground point in the pixel: the mapped point is determined to be an invalid point; The pixel is a ground pixel or a ground repair pixel and the mapped point only occludes ground points: the mapped point is determined to be a dynamic point; The pixel is a ground pixel or a ground repair pixel and the mapped point both occludes ground points and is occluded by a static point: the mapped point is determined to be a suspected dynamic point; After the occlusion detection of the current frame elevation image is completed, all mapped points are determined to be one of the following five types: static point, dynamic point, suspected dynamic point, suspected static point, or invalid point; Integrate the occlusion detection results of the elevation images of the previous M frames. If the number of times a certain mapped point is determined to be the same type in five frames exceeds (M + 1) / 2 times, it is finally marked as that type; otherwise, it is marked as a static point.

6. The method for dynamic point cloud detection based on elevation images as claimed in claim 1, wherein performing a regional consistency check on the pixels where the dynamic points, the suspected dynamic points, and the suspected static points are located to determine the final dynamic points, including: Define the neighborhood of the current pixel as the 8-pixel area around it, and determine whether there is a point that meets the conditions among the neighborhood pixels. If the Z-axis height difference between a certain type of point in the neighborhood pixels and the marked point is less than the set threshold , then modify the type of the marked point according to specific rules, including: When the marked point is a dynamic point, if there are static points that meet the conditions among the neighboring pixels, and the number of times this condition is met exceeds three times in five frames of elevation images, then the marked point is re-determined as a static point; otherwise, the marked point is still determined as a dynamic point; When the marked point is a suspected dynamic point, if there are static points that meet the conditions among the neighboring pixels, and the number of times this condition is met exceeds two times in five frames of elevation images, then the marked point is re-determined as a static point; otherwise, the marked point is determined as a dynamic point; When the marked point is a suspected static point, if there are static points that meet the conditions among the neighboring pixels, and the number of times this condition is met exceeds one time in five frames of elevation images, then the marked point is re-determined as a static point; otherwise, the marked point is determined as a dynamic point.

7. The method for dynamic point cloud detection based on elevation images as claimed in claim 1 or 2, wherein for the point cloud determined to be a dynamic point after the regional consistency check, performing a clustering operation using a density-based spatial clustering algorithm with noise.

8. A dynamic point cloud detection system based on elevation images, characterized in that, Including: a data acquisition unit configured to: acquire the ground point cloud, the non-ground point cloud, and the odometer information of each frame of lidar point cloud; an elevation image construction unit configured to: construct an elevation image using the static point cloud and the ground point cloud obtained from the previous frame, including: converting each point in the static point cloud and the ground point cloud from the Cartesian coordinate system to the polar coordinate system, and mapping each point to the corresponding elevation image pixel according to the resolution of the elevation image, where the index of each pixel consists of a ring area number and a sector number, and the elevation image includes four circular regions, namely circular region one, circular region two, circular region three, and circular region four, and the pixel density gradually decreases from the inside to the outside; a ground continuity repair unit configured to: repair the ground continuity of the elevation image; an occlusion detection unit configured to: combine the odometer information and perform occlusion detection on the point cloud of the current frame based on the elevation image to determine dynamic points, suspected dynamic points, and suspected static points; a regional consistency check unit configured to: perform a regional consistency check on the pixels where the dynamic points, the suspected dynamic points, and the suspected static points are located to determine the final dynamic points; a detection result generation unit configured to: perform a clustering operation on the detected final dynamic point cloud to obtain the dynamic point cloud detection result; Among them, in the occlusion detection unit, combining the odometer information and performing occlusion detection on the point cloud of the current frame based on the elevation image to determine dynamic points, suspected dynamic points, and suspected static points, including: always maintaining the elevation images of the most recent M frames of lidar point cloud. In the initialization stage, a time window of the first five frames of lidar point cloud is required. During this period, all non-ground point clouds are regarded as static point clouds, and the point clouds are only used to construct the elevation image without performing occlusion detection. Starting from the (M + 1)-th frame, occlusion detection is performed on the point cloud; Using odometry data, the non-ground point clouds of the current frame are transformed one by one into the coordinate system of the elevation images of the previous five frames; The transformed point clouds are divided into corresponding pixels according to the resolution of the corresponding elevation images, including: first converting the points from the Cartesian coordinate system to the polar coordinate system, and then mapping each point to the corresponding pixel according to the resolution of the elevation image; Perform occlusion detection on the pixel where the point to be detected mapped in the elevation image. If the difference between the Z-axis height of the point to be detected and the Z-axis height of the highest ground point in the pixel exceeds the set threshold , it is considered that the point to be detected occludes the ground point; otherwise, it is considered that no occlusion occurs. If the difference between the Z-axis height of the static point in the pixel and the Z-axis height of the point to be detected exceeds the threshold , it is considered that the point to be detected is occluded by the static point; otherwise, it is considered that no occlusion occurs.

Images