A tire point cloud segmentation method

Abstract

The present invention discloses a tire point cloud segmentation method, step 1: record the point cloud set containing the tire as point, calculate all points P in the set i The corresponding curvature value Curv i and the corresponding normal vector #imgabs0#Step 2: Select the contact point P between the tire and the ground touch Add the seed point set Seed as the initial seed point and define a tire area set #imgabs1# where #imgabs2# represents an empty set; Step 3: Take a seed point P from Seed S , and P S Delete from the set Seed, and traverse all neighboring points P in the sphere with the seed point as the center and r as the radius j , j=1,2,…,M, and the neighborhood points P that meet the region growth conditions j Add the neighboring point P that meets the seed point selection conditions to the set tire j Add the set seed; Step 4: Enter step 3 loop until #imgabs3# At this time, all the point clouds in tire are the tire point clouds obtained by region growing; the present invention can specifically select the contact point of the target tire to be segmented as the seed point, and will not generate time-consuming point cloud segmentation in non-target areas.

Description

Technical Field

[0001] The present invention relates to the field of computer vision three-dimensional measurement and intelligent transportation technology, and in particular to a tire point cloud segmentation method that detects tire contact points and then uses the tire contact points as seed points for single-region growth. Background Art

[0002] With the continuous development of computer vision 3D measurement, it is now being used in the intelligent transportation sector to achieve non-contact measurement of vehicle dimensions and wheelbase. Tire point cloud segmentation is crucial for wheelbase measurement. The segmented tire point cloud is used to fit the tire center point and calculate the wheelbase. Therefore, tire point cloud segmentation is a crucial step in wheelbase measurement.

[0003] Currently, many scholars have proposed different point cloud segmentation methods, among which the technical solutions that are closer to the present invention are: Reference (Yang Na, Qin Zhiyuan, et al. Research on airborne LIDAR point cloud segmentation method for ground point recognition [J]. Surveying and Mapping Engineering, 2014, 23 (10): 18-22.) proposed to determine the seed points and seed planes for regional growth based on the normal vector and residual of each laser foot point, and in the process of regional growth, the distance from the adjacent point to the seed plane and the angle difference between the adjacent point and the seed point are used as similarity measurement standards. The seed point determination method of this method will produce seed points in non-segmented target areas, resulting in regional growth occurring in non-segmented target areas, increasing the time consumption of point cloud segmentation; Reference (Liu Qingqun, Li Hao, et al. 3DHough transform point cloud segmentation method for rock mass structural surface extraction [J]. Engineering Survey, 2017 (4): 64-67.) proposed to randomly select seed points and combine 3D Hough transform is used to extract the seed surface, and then all points on the seed surface are used as seed points for region growth. During the region growth process, the angle between the normal vectors of the adjacent points and the seed points is used as a constraint condition. This method randomly selects seed points, which is more likely to cause the seed points to fall in the non-segmented target area, resulting in time-consuming region growth of the point cloud in the non-segmented target area. Moreover, only the angle between the normal vectors of the adjacent points and the seed points is used as a constraint condition for region growth, which easily leads to over-segmentation or under-segmentation of region growth. The literature (Liu Wei, Li Yan, et al. An improved region growth point cloud segmentation method under unstructured environment [J]. Science, Technology and Engineering, 2021, 21(27): 11650-11655.) proposes to use the point with the smallest curvature as the seed point and set the spatial threshold range for region growth. The seed point selection method of this method will also cause the seed point to fall in the non-segmented target area, resulting in region growth occurring in the non-segmented target area, which increases time consumption.

[0004] In summary, the current region growing point cloud segmentation method has the following shortcomings: (1) it cannot select the seed points of the target area to be segmented in a targeted manner, which increases the segmentation time; (2) the constraint conditions of region growing are single, which easily causes overgrowth or undergrowth. Summary of the Invention

[0005] In response to the above problems existing in the existing region growing point cloud segmentation method, the present invention proposes a tire point cloud segmentation method that detects the tire contact point and then uses the tire contact point as a seed point to perform single region growth.

[0006] The technical solutions of the present invention are as follows:

[0007] A tire point cloud segmentation method comprises the following steps:

[0008] Step 1: Record the point cloud set containing the tire as point and calculate all points P in the set i The corresponding curvature value Curv i and the corresponding normal vector Where i represents the i-th point cloud in point, N represents the number of point clouds in point, i = 1, 2, ..., N;

[0009] Step 2: Select the contact point P between the tire and the ground touch Add the seed point set Seed as the initial seed point and define a tire area set in Represents an empty set;

[0010] Step 3: Take a seed point P from Seed S , and P S Delete from the set Seed, and traverse all neighboring points P in the sphere with the seed point as the center and r as the radius j , j=1,2,…,M, and the neighborhood points P that meet the region growth conditions j Add the neighboring point P that meets the seed point selection conditions to the set tire j Add to the set seed, where j represents the seed point P S The jth neighborhood point of M represents the seed point P S The number of neighboring points of ;

[0011] Step 4: Enter step 3 loop until At this time, all the point clouds in tire are the tire point clouds obtained by region growing.

[0012] Furthermore, the contact point P between the tire and the ground in step 2 touch The selections are as follows:

[0013] 2.1): Divide the area from the ground to a height less than h into h×h×h voxels, denoted as voxel k , k=1,2,…,K, where voxel k represents the kth voxel, K represents the total number of voxels;

[0014] 2.2): Record each voxel k The number of point clouds in , recorded as count k ;

[0015] 2.3): Find more than n continuous voxels in space, and the number of point clouds in each voxel is greater than thresh count , thresh count Represents the pre-set threshold value of the number of point clouds in a voxel;

[0016] 2.4): Calculate the middle point of the continuous voxel found in step 2.3) as the contact point P between the tire and the ground touch .

[0017] Furthermore, the neighborhood point P in step 3 j The conditions for region growing are as follows:

[0018] 3.1): Calculate the neighborhood point P according to formula (1) j and seed point P S The angle θ between the normal vectors Sj ;

[0019]

[0020] 3.2): If θ Sj ≤θ threshold , then point P j Satisfy the conditions for regional growth;

[0021] in, Represents the seed point P S The corresponding normal vector, Represents the neighborhood point P j The corresponding normal vector, θ threshold Indicates the preset normal vector angle threshold.

[0022] Furthermore, the neighborhood point P in step 3 j The conditions for seed point selection must meet the following two conditions at the same time: (1) The neighboring point P j Satisfy the region growth conditions determined in step 3.2); (2) Curv j <C threshold ; Among them, Curv j Represents the neighborhood point P jThe corresponding curvature value, C threshold Indicates the preset curvature value threshold.

[0023] The beneficial effects of the present invention are as follows: by using the method of the present invention to perform point cloud segmentation on a tire, it is possible to specifically select the contact points of the target tire to be segmented as seed points, without generating time-consuming point cloud segmentation of non-target areas, and the constraints of region growth include the normal vector angle between the neighboring point and the seed point and the curvature value of the neighboring point, which is less likely to cause overgrowth and undergrowth. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is the point cloud set point selected in the embodiment of the present invention;

[0025] Figure 2 The contact point P between the tire and the ground selected in the embodiment of the present invention is touch Schematic diagram of the location;

[0026] Figure 3 This is the result of tire segmentation using region growing in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The following is a detailed description of the tire point cloud segmentation method according to the present invention with reference to the following embodiments. It should be understood that the embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0028] A tire point cloud segmentation method of the present invention comprises the following steps:

[0029] Step 1: Record the point cloud set containing the tire as point and calculate all points P in the set i The corresponding curvature value Curv i and the corresponding normal vector Where i represents the i-th point cloud in point, N represents the number of point clouds in point, i = 1, 2, ..., N, and the point cloud set point selected in this embodiment is as follows Figure 1 As shown;

[0030] Step 2: Select the contact point P between the tire and the ground touch Add the seed point set Seed as the initial seed point and define a tire area set in represents an empty set. In this embodiment, the contact point P between the tire and the ground is selected touch like Figure 2 As shown;

[0031] The contact point P between the tire and the ground in step 2 touch The selections are as follows:

[0032] 2.1): Divide the area from the ground to a height less than h into h×h×h voxels, denoted as voxel k , k=1,2,…,K, where voxel k represents the kth voxel, K represents the total number of voxels;

[0033] 2.2): Record each voxel k The number of point clouds in , recorded as count k ;

[0034] 2.3): Find more than n continuous voxels in space, and the number of point clouds in each voxel is greater than thresh count , thresh count Represents the pre-set threshold value of the number of point clouds in a voxel;

[0035] 2.4): Calculate the middle point of the continuous voxel found in step 2.3) as the contact point P between the tire and the ground touch .

[0036] In this embodiment, the height h is set to 30 mm, n is set to 4, and thresh count Set to 3.

[0037] Step 3: Take a seed point P from Seed S , and P S Delete from the set Seed, and traverse all neighboring points P in the sphere with the seed point as the center and r as the radius j , j=1,2,…,M, and the neighborhood points P that meet the region growth conditions j Add the neighboring point P that meets the seed point selection conditions to the set tire j Add to the set seed, where j represents the seed point P S The jth neighborhood point of M represents the seed point P S The number of neighborhood points, in this embodiment, the sphere radius r is set to 30 mm;

[0038] Neighborhood point P in step 3 j The conditions for region growing are as follows:

[0039] 3.1): Calculate the neighborhood point P according to formula (1) j and seed point P S The angle θ between the normal vectors Sj ;

[0040]

[0041] 3.2): If θ Sj ≤θthreshold , then point P j Satisfy the conditions for regional growth;

[0042] in, Represents the seed point P S The corresponding normal vector, Represents the neighborhood point P j The corresponding normal vector, θ threshold represents the preset normal vector angle threshold, in this embodiment θ threshold Set to 3.5°.

[0043] Neighborhood point P in step 3 j The conditions for seed point selection must meet the following two conditions at the same time:

[0044] (1) Neighborhood point P j Satisfy the region growth conditions determined in step 3.2); (2) Curv j <C threshold ; Among them, Curv j Represents the neighborhood point P j The corresponding curvature value, C threshold represents a preset curvature value threshold. In this embodiment, C threshold Set to 0.1.

[0045] In this embodiment, through the above processing, it can be seen that Figure 3 The tires are accurately segmented.

[0046] Step 4: Enter step 3 loop until At this time, all the point clouds in tire are tire point clouds obtained by region growing. In this embodiment, the tire point cloud obtained by region growing is as follows: Figure 3 shown.

Claims

1. A tire point cloud segmentation method, characterized in that: The steps include: Step 1: Record the point cloud set containing the tire as point and calculate all points P in the set i The corresponding curvature value Curv i and the corresponding normal vector Where i represents the i-th point cloud in point, N represents the number of point clouds in point, i = 1, 2, ..., N; Step 2: Select the contact point P between the tire and the ground touch Add the seed point set Seed as the initial seed point and define a tire area set in Represents an empty set; Step 3: Take a seed point P from Seed S , and P S Delete from the set Seed, and traverse all neighboring points P in the sphere with the seed point as the center and r as the radius j , j=1,2,…,M, and the neighborhood points P that meet the region growth conditions j Add the neighboring point P that meets the seed point selection conditions to the set tire j Add to the set seed, where j represents the seed point P S The jth neighborhood point of M represents the seed point P S The number of neighboring points of ; Neighborhood point P in step 3 j The conditions for region growing are as follows: 3.1): Calculate the neighborhood point P according to formula (1) j and seed point P S The angle θ between the normal vectors Sj ; 3.2): If θ Sj ≤θ threshold , then point P j Satisfy the conditions for regional growth; in, Represents the seed point P S The corresponding normal vector, Represents the neighborhood point P j The corresponding normal vector, θ threshold Indicates the preset normal vector angle threshold; Step 4: Enter step 3 loop until At this time, all the point clouds in tire are the tire point clouds obtained by region growing.

2. The tire point cloud segmentation method according to claim 1, characterized in that: The contact point P between the tire and the ground in step 2 touch The selections are as follows: 2.1): Divide the area from the ground to a height less than h into h×h×h voxels, denoted as voxel k , k=1,2,…,K, where voxel k represents the kth voxel, K represents the total number of voxels; 2.2): Record each voxel k The number of point clouds in , recorded as count k ; 2.3): Find more than n continuous voxels in space, and the number of point clouds in each voxel is greater than thresh count , thresh count Represents the pre-set threshold value of the number of point clouds in a voxel; 2.4): Calculate the middle point of the continuous voxel found in step 2.3) as the contact point P between the tire and the ground touch .

3. The tire point cloud segmentation method according to claim 1, characterized in that: Neighborhood point P in step 3 j The conditions for seed point selection must meet the following two conditions at the same time: 1) Neighborhood point P j Satisfy the region growth conditions determined in step 3.2); 2)Curv j <C threshold ; Among them, Curv j Represents the neighborhood point P j The corresponding curvature value, C threshold Indicates the preset curvature value threshold.

Images