A method and system for object detection in lidar point clouds

Abstract

The application discloses a laser radar point cloud object detection method and system, comprising the following steps: obtaining a three-dimensional point cloud of a laser radar, and after filtering interference point clouds and compensating for missing point clouds, obtaining a to-be-segmented point cloud; generating a two-dimensional distance image from the to-be-segmented point cloud; setting an angle constraint; the angle constraint is that the included angle between the line connecting two adjacent points in the two-dimensional distance image and the line connecting the radar center and the point farther away from the radar center among the two adjacent points is not less than an angle threshold; setting a distance constraint; the distance constraint is that the distance between the two adjacent points in the two-dimensional distance image is not greater than a distance threshold; traversing the points in the two-dimensional distance image, judging whether the two adjacent points satisfy the angle constraint or the distance constraint, if any constraint is satisfied, marking the two adjacent points as the same object, otherwise, marking the two adjacent points as different objects, and thus obtaining an object segmentation result. The object can be quickly detected under sparse point cloud data.

Description

A method and system for object detection in lidar point clouds Technical Field

[0001] This invention relates to the field of point cloud segmentation technology, and in particular to a method and system for detecting objects in point clouds using lidar. Background Technology

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

[0003] Mobile robot technology can be applied to fields such as inspection and maintenance, logistics and transportation, and autonomous driving. While robots demonstrate sufficient reliability in experimental scenarios, they encounter numerous problems when faced with complex real-world applications. As the "eyes" of a mobile robot, the environmental perception module directly impacts its ability to find a safe path. Commonly used sensors include LiDAR, RGB cameras, and millimeter-wave radar.

[0004] LiDAR, a widely used sensor, utilizes rotating laser beams to reflect and acquire information about the surrounding environment, outputting the information as a dense 3D point cloud. It boasts advantages such as high accuracy, large range, and minimal susceptibility to lighting conditions. However, while high-beam LiDAR offers higher accuracy, it also incurs higher costs, while low-beam LiDAR suffers from sparse point clouds. Therefore, reducing costs while maintaining object recognition accuracy presents a challenge for processing and segmenting point clouds.

[0005] In addition, point cloud segmentation is performed using only angle constraints in most scenarios. However, when dealing with continuous large-area objects, such as walls, a single angle parameter cannot determine whether the small angle is due to the depth change between the two object points or the laser beam being almost parallel at a distance on the wall. Therefore, segmentation failures may occur. Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes a lidar point cloud object detection method and system. The method segments the point cloud based on angle and distance constraints, avoiding the failure of single-angle segmentation methods at the edges of continuous objects. This enables rapid object detection in sparse point cloud data, meeting the real-time requirements for practical applications.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] In a first aspect, the present invention provides a method for detecting objects in lidar point clouds, comprising:

[0009] After acquiring the 3D point cloud of the lidar and filtering out the interfering point cloud and compensating for the missing point cloud, the point cloud to be segmented is obtained, and a 2D distance image is generated from the point cloud to be segmented.

[0010] Set an angle constraint; the angle constraint is that the angle formed by the line connecting two adjacent points in the two-dimensional distance image and the radar center and the line connecting the two adjacent points that are farther from the radar center is not less than the angle threshold.

[0011] Set a distance constraint; the distance constraint is that the distance between two adjacent points in the two-dimensional distance image is not greater than a distance threshold.

[0012] Traverse the points in the 2D distance image and determine whether two adjacent points satisfy the angle constraint or the distance constraint. If either constraint is satisfied, mark the two adjacent points as the same object; otherwise, mark them as different objects. This yields the object segmentation result.

[0013] As an alternative implementation, the process of filtering out interfering point clouds includes screening the lidar 3D point cloud based on reflection intensity and incident angle.

[0014] As an alternative implementation method, the filtering process for interfering point clouds also includes filtering out ground point clouds, specifically:

[0015] The 3D point cloud is mapped to a 2D point cloud and divided into N equal regions. Each region is further divided into M blocks. A ground curve is fitted to each region.

[0016] The point with the lowest height in each block of the region is selected as the initial point. The vertical projection of the remaining points onto the ground curve is used as the criterion for judgment. This is used to classify and obtain ground point clouds and object point clouds, and object point clouds are retained.

[0017] As an alternative implementation method, the compensation for missing point clouds involves determining the range of the missing point clouds and performing linear interpolation compensation.

[0018] As an alternative implementation, the included angle θ is:

[0019]

[0020] Where 0 < θ < 90°, d1 is the distance from the radar center to the point farther from the radar center between two adjacent points, and d2 is the distance from the other point to the radar center. The angle between adjacent laser beams.

[0021] As an alternative implementation, the included angle threshold is less than or equal to 10°.

[0022] As an alternative implementation, the distance threshold is based on the distance d1 from one of the two adjacent points to the radar center and the angle between the adjacent laser beams. We obtain the length of the waist with d1 as the waist length. Given the length of the base of the triangle at the vertex, multiply the base length by the first correction coefficient to obtain the standard distance between two adjacent points, and multiply the standard distance by the second correction coefficient to obtain the distance threshold.

[0023] In a second aspect, the present invention provides a lidar point cloud object detection system, comprising:

[0024] The preprocessing module is configured to acquire the 3D point cloud of the LiDAR, and after filtering out the interfering point cloud and compensating for the missing point cloud, obtain the point cloud to be segmented, and generate a 2D distance image from the point cloud to be segmented.

[0025] An angle constraint setting module is configured to set angle constraints; the angle constraint is that the angle formed by the line connecting two adjacent points in the two-dimensional distance image and the radar center and the line connecting the two adjacent points that are farther from the radar center is not less than the angle threshold.

[0026] The distance constraint setting module is configured to set distance constraints; the distance constraint is that the distance between two adjacent points in a two-dimensional distance image is not greater than a distance threshold.

[0027] The object segmentation module is configured to traverse the points in the two-dimensional distance image, determine whether two adjacent points satisfy the angle constraint or the distance constraint, and mark the two adjacent points as the same object if either constraint is satisfied, otherwise they are different objects, thus obtaining the object segmentation result.

[0028] Thirdly, the present invention provides an electronic device including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0029] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in the first aspect.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0031] This invention proposes a method and system for object detection using lidar point clouds. It separates ground point clouds from object point clouds by segmenting and fitting the ground, filters out distorted points using physical information such as the laser intensity returned by the radar, compensates for missing object point clouds using linear interpolation based on the point cloud scanning sequence, and segments the point cloud based on angle and distance constraints. This avoids the failure of single-angle segmentation methods at the edges of continuous objects, enabling rapid object detection in sparse point cloud data. The system operates quickly, meeting the real-time requirements of practical applications, and achieves higher clustering accuracy and better detail in sparse point cloud environments.

[0032] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0033] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0034] Figure 1 is a flowchart of the lidar point cloud object detection method provided in Embodiment 1 of the present invention;

[0035] Figure 2 is a schematic diagram of ground curve fitting provided in Embodiment 1 of the present invention;

[0036] Figures 3(a) and 3(b) show the angle segmentation principle and wall segmentation failure analysis diagram provided in Embodiment 1 of the present invention;

[0037] Figure 4 is a statistical chart of the algorithm running speed provided in Embodiment 1 of the present invention;

[0038] Figure 5 is a ground segmentation effect diagram provided in Embodiment 1 of the present invention;

[0039] Figures 6(a) and 6(b) are comparison images of wall point cloud segmentation provided in Embodiment 1 of the present invention;

[0040] Figures 7(a) and 7(b) are comparison images of parking lot vehicle point cloud segmentation provided in Embodiment 1 of the present invention. Detailed Implementation

[0041] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0042] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0043] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. Furthermore, it should be understood that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0044] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0045] Example 1

[0046] This embodiment provides a method for object detection in lidar point clouds, as shown in Figure 1, including:

[0047] After acquiring the 3D point cloud of the lidar and filtering out the interfering point cloud and compensating for the missing point cloud, the point cloud to be segmented is obtained, and a 2D distance image is generated from the point cloud to be segmented.

[0048] Set an angle constraint; the angle constraint is that the angle formed by the line connecting two adjacent points in the two-dimensional distance image and the radar center and the line connecting the two adjacent points that are farther from the radar center is not less than the angle threshold.

[0049] Set a distance constraint; the distance constraint is that the distance between two adjacent points in the two-dimensional distance image is not greater than a distance threshold.

[0050] Traverse the points in the 2D distance image and determine whether two adjacent points satisfy the angle constraint or the distance constraint. If either constraint is satisfied, mark the two adjacent points as the same object; otherwise, mark them as different objects. This yields the object segmentation result.

[0051] The 3D point cloud data directly transmitted by LiDAR often contains points with distorted positions or low quality, which can interfere with subsequent segmentation and clustering. Therefore, these points need to be filtered out. In point cloud segmentation and clustering, ground point clouds are also a serious source of interference. Therefore, the filtering of interfering point clouds in this embodiment includes the filtering out of distorted or low-quality points, as well as the filtering out of ground point clouds.

[0052] In this embodiment, the filtering process for distorted or low-quality points includes:

[0053] Incident intensity and incident angle are used as the basis for measuring point cloud quality;

[0054] Let P be a point in a 3D point cloud, and let the incident intensity I be I(p) = R / D(P); where R is the reflectivity of the object, measured by the lidar sensor, and D(P) is the distance from the point to the center of the lidar coordinates; or the incident intensity can also be directly given by the lidar sensor, such as the Velodyne lidar.

[0055] Finally, points with excessively high or low incident intensity, or incident angles close to 180° or 0°, are filtered out.

[0056] In this embodiment, the filtering process of the ground point cloud includes: using a segmented sampling method to fit the three-dimensional point cloud to the ground curve, as shown in Figure 2, and classifying it into ground points and non-ground points, retaining the non-ground points; and in order to improve computational efficiency, the three-dimensional point cloud (x,y,z) is mapped to the two-dimensional point cloud (d,z) for processing, avoiding the complex neighborhood relationship establishment and feature extraction.

[0057] Specifically: Treat the ground as a circle with an infinite radius, and divide the two-dimensional point cloud into N equal regions S, with each region having an angle Δα.

[0058] Each region is further divided into M equal blocks b, and a ground curve z = kd + b is fitted to the region;

[0059] Find the point with the lowest height z in each block of the region as the initial point, and use the vertical projection of the remaining points onto the ground curve as the criterion. Based on the fitted ground curve, determine whether the ground point cloud or the object point cloud is a point cloud; points greater than the threshold are non-ground points, and points less than the threshold are ground points.

[0060] After completion, proceed to the next area to continue calculations until the classification of all point clouds is completed.

[0061] In practical applications, objects appearing near the radar may be mistaken for ground starting points, leading to subsequent classification errors. Therefore, the minimum height of block b within the near-range is set to the radar installation height to avoid this problem and speed up the calculation. Due to the piecewise sampling and fitting of the ground curve, the ground point cloud can be correctly segmented even on undulating terrain.

[0062] In this embodiment, considering that when a laser beam illuminates an object such as a car window, not all of the laser beam returns to the radar sensor, resulting in missing point cloud data for that portion, this can lead to incomplete object recognition and oversegmentation in sparse point cloud scenarios. Therefore, the point cloud data returned by the radar is traversed in each frame. When there is a certain number S missing points between two points with data, and the two points satisfy a certain distance D = |d p1 -d p2 If the data is in the middle, it is determined that the midpoint is missing data; assuming that these data approximately satisfy a linear relationship, the missing data is filled in using linear interpolation:

[0063] In this embodiment, the 3D point cloud measured by the lidar is represented using a spherical coordinate system. The coordinates of point P in the spherical coordinate system are represented as P = [L, β, α, ...]. T L, β, and α represent the distance, elevation, and azimuth components of point P, respectively.

[0064]

[0065] Therefore, the processed point cloud to be segmented is converted into a two-dimensional distance image in spherical coordinates based on the coordinate information.

[0066] In this embodiment, by setting angle constraints and distance constraints, the points in the two-dimensional distance image are traversed to determine whether two adjacent points satisfy the angle constraints and distance constraints. If either constraint is satisfied, the two adjacent points are marked as the same object. As shown in Figure 1, if the angle constraint is satisfied, it means that they are the same object. If the angle constraint is not satisfied but the distance constraint is satisfied, it also means that they are the same object. If neither is satisfied, they are different objects, until all objects are identified.

[0067] Figure 3(a) shows two points A and B measured by two adjacent laser beams. O represents the radar center in spherical coordinates. d1 is the distance from point A to the radar center O, i.e., OA, and d2 is the distance from point B to the radar center O, i.e., OB. The angle between A and B is defined as the angle θ between the straight line AB and the line OA connecting the radar center and the point farther from the radar center, and 0 < θ < 90°.

[0068]

[0069] in, It is the angle between adjacent laser beams, a known angle that can be obtained by consulting the lidar user manual;

[0070] Let the included angle threshold be θ0. The included angle threshold θ0 can help identify whether two adjacent points come from two independent objects or from the same object in clustering.

[0071] If θ is less than θ0, then the difference in depth between two adjacent points is too large, and they belong to two different objects; otherwise, the two adjacent points are temporarily considered to belong to the same object.

[0072] In this embodiment, the selection of the included angle threshold θ0 is closely related to the distance between objects in the environment. Generally, to ensure the correctness of the segmentation, the included angle threshold θ0 is less than or equal to 10°.

[0073] In most scenarios, good segmentation results can be achieved using only angle constraints. However, in special cases, such as when dealing with continuous large-area objects like walls, segmentation failures may occur. As shown in Figure 3(b), when the laser beam scans the wall away from the radar, the angle θ continuously decreases, eventually falling below the set angle threshold at a distant point on the wall. Consequently, the wall surface is segmented into multiple objects. This is because a single angle parameter cannot determine whether the excessively small angle θ originates from depth variations between two object points or from the laser beam being almost parallel at a distant point on the wall.

[0074] To avoid segmentation failures on large, continuous wall surfaces, this embodiment adds distance constraints. Specifically, the distance L between two adjacent points is calculated. When the distance L is greater than or equal to a distance threshold L0, the points are determined to be on the same continuous object; otherwise, they are on two separate objects.

[0075] The distance threshold L0 is set as follows:

[0076] Analysis of the laser beam in Figure 3(b) reveals that the angle along a certain direction on the wall is continuous, and the length d2 of the previous angle calculation is the same as the length d1 of the current angle calculation. Therefore, the angle between d1 and... Given the given information, find the length of the leg with d1 as its length. Let L be the length of the base of the triangle at the vertex; multiply the base length by a correction factor k1 to approximate the standard distance L between two adjacent points. ′ 0; Considering that continuous large objects can have curves in addition to flat walls, the standard distance L is set to 0. ‘ Multiplying 0 by the correction coefficient k2 yields the distance threshold L0;

[0077]

[0078] The correction factor can be adjusted according to the curvature of the wall in a specific scenario.

[0079] To verify the effectiveness of the method in this embodiment, comparative experiments were conducted with angular clustering and Euclidean clustering methods in the same scenario. The platform on which the lidar was installed was an autonomous vehicle equipped with various sensors. A Velodyne 16-line mechanical lidar was used as a scanner to collect point clouds. The lidar was installed at the front left of the vehicle, with a horizontal field of view of 270° and a vertical field of view of ±15°.

[0080] First, 50 consecutive frames of point cloud segmentation time were collected at the test site. Calculations showed that the average segmentation time per frame was 2.597 ms, and the maximum was 2.901 ms, both less than 5 ms. This demonstrates that the method in this embodiment can run quickly on a robot processor, meeting the requirement of real-time online processing of radar point clouds. The standard deviation was 0.102, indicating that the method in this embodiment has good stability, as shown in Figure 4.

[0081] Then, the segmentation of the ground point cloud was tested. As shown in Figure 5, when facing a sloping ground, the method of this embodiment can correctly and completely identify the ground point cloud and classify it for subsequent calculation.

[0082] When dealing with large, continuous objects, such as continuous walls, the single angle segmentation algorithm exhibits oversegmentation at a distance from the wall, resulting in the wall being divided into multiple objects, as shown in Figure 6(a). However, the method in this embodiment, after adding distance constraints, correctly segments the continuous wall, as shown in Figure 6(b).

[0083] In a static parking lot environment, the method in this embodiment achieved complete and accurate identification of all five vehicles and one pedestrian, as shown in Figure 7(b); while in the comparison, Euclidean clustering only correctly identified one vehicle and one pedestrian, and exhibited oversegmentation (the second vehicle from the right was only half identified), undersegmentation (the second and third vehicles from the left were identified as the same object), and non-identification (the first vehicle from the right), as shown in Figure 7(a).

[0084] To further test the algorithm's performance, a total of 900 frames (90 seconds) of data were collected in the scene. The ratio of the number of correctly detected frames to the total number of frames was defined as the detection rate. The statistical results are shown in Table 1. It can be seen that the algorithm in this embodiment has a higher detection rate and a faster segmentation speed.

[0085] Table 1. Algorithm Comparison Results in Parking Lot Scenarios

[0086] Total Frames | Correct Frame Detection Rate | Average Segmentation Time (ms) | Euclidean Segmentation Method | 900686 | 0.76236.00 | Angle Segmentation Method | 900827 | 0.919 | 2.083 | Algorithm of this Example | 900831 | 0.923 | 2.597 surface

[0087] Example 2

[0088] This embodiment provides a lidar point cloud object detection system, including:

[0089] The preprocessing module is configured to acquire the 3D point cloud of the LiDAR, and after filtering out the interfering point cloud and compensating for the missing point cloud, obtain the point cloud to be segmented, and generate a 2D distance image from the point cloud to be segmented.

[0090] An angle constraint setting module is configured to set angle constraints; the angle constraint is that the angle formed by the line connecting two adjacent points in the two-dimensional distance image and the radar center and the line connecting the two adjacent points that are farther from the radar center is not less than the angle threshold.

[0091] The distance constraint setting module is configured to set distance constraints; the distance constraint is that the distance between two adjacent points in a two-dimensional distance image is not greater than a distance threshold.

[0092] The object segmentation module is configured to traverse the points in the two-dimensional distance image, determine whether two adjacent points satisfy the angle constraint or the distance constraint, and mark the two adjacent points as the same object if either constraint is satisfied, otherwise they are different objects, thus obtaining the object segmentation result.

[0093] It should be noted that the above modules correspond to the steps described in Embodiment 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should also be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0094] In further embodiments, the following is also provided:

[0095] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0096] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0097] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0098] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in Embodiment 1.

[0099] The method in Example 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0100] Those skilled in the art will recognize that the units, i.e., algorithm steps, of the various examples described in connection with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0101] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for object detection in lidar point clouds, characterized in that, include: After acquiring the 3D point cloud of the lidar and filtering out the interfering point cloud and compensating for the missing point cloud, the point cloud to be segmented is obtained, and a 2D distance image is generated from the point cloud to be segmented. Angle constraints are set; the angle constraint is that the angle formed by the line connecting two adjacent points in the two-dimensional distance image and the radar center, and the line connecting the two adjacent points that are farther from the radar center, is not less than the included angle threshold; the included angle for: Where, 0 < <90°， This represents the distance from the radar center to the point farther from the radar center between two adjacent points. This is the distance from another point to the radar center. The angle between adjacent laser beams is defined; the angle threshold is less than or equal to 10°; a distance constraint is set; the distance constraint is that the distance between two adjacent points in the two-dimensional distance image is not greater than a distance threshold; the distance threshold is determined based on the distance from one of the two adjacent points to the radar center. and the angle between adjacent laser beams , to obtain For waist length, Given the length of the base of the triangle at the vertex, multiply the base length by the first correction coefficient to obtain the standard distance between two adjacent points. Multiply the standard distance by the second correction coefficient to obtain the distance threshold. Traverse the points in the two-dimensional distance image and determine whether two adjacent points satisfy the angle constraint or the distance constraint. If either constraint is satisfied, mark the two adjacent points as the same object; otherwise, mark them as different objects. This yields the object segmentation result.

2. The lidar point cloud object detection method as described in claim 1, characterized in that, The process of filtering out interfering point clouds includes screening the lidar 3D point cloud based on reflection intensity and incident angle.

3. The lidar point cloud object detection method as described in claim 1, characterized in that, The process of filtering out interfering point clouds also includes filtering out ground point clouds. Specifically, the three-dimensional point cloud is mapped to a two-dimensional point cloud and divided into N regions. Each region is further divided into M blocks. A ground curve is fitted to each region. The point with the lowest height in each block of the region is selected as the initial point. The vertical projection of the remaining points onto the ground curve is used as the criterion. This process is used to classify ground point clouds and object point clouds, and the object point clouds are retained.

4. The lidar point cloud object detection method as described in claim 1, characterized in that, The compensation for missing point clouds involves determining the range of the missing point clouds and performing linear interpolation compensation.

5. A lidar point cloud object detection system, characterized in that, include: The preprocessing module is configured to acquire a 3D point cloud from the lidar, filter out interfering point clouds, and compensate for missing point clouds to obtain the point cloud to be segmented, generating a 2D distance image from the point cloud to be segmented. The angle constraint setting module is configured to set angle constraints; the angle constraint is that the angle formed by the line connecting two adjacent points in the 2D distance image and the radar center, and the line connecting the two adjacent points with the point farther from the radar center, is not less than an angle threshold. for: Where, 0 < <90°， This represents the distance from the radar center to the point farther from the radar center between two adjacent points. This is the distance from another point to the radar center. The angle between adjacent laser beams; the angle threshold is less than or equal to 10°; the distance constraint setting module is configured to set a distance constraint; the distance constraint is that the distance between two adjacent points in the two-dimensional distance image is not greater than a distance threshold; the distance threshold is: based on the distance from one of the two adjacent points to the radar center. and the angle between adjacent laser beams , to obtain For waist length, The base length of the triangle at the vertex is given. Multiplying the base length by the first correction coefficient yields the standard distance between two adjacent points. Multiplying the standard distance by the second correction coefficient yields the distance threshold. The object segmentation module is configured to traverse the points in the two-dimensional distance image and determine whether two adjacent points satisfy the angle constraint or the distance constraint. If either constraint is satisfied, the two adjacent points are marked as the same object; otherwise, they are different objects, thus obtaining the object segmentation result.

6. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method according to any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, perform the method described in any one of claims 1-4.

Images