A method for identifying and positioning the grasping points of an insulator detection tool based on the point cloud of a solid-state lidar

Abstract

The present invention discloses a method for identifying and positioning the grasping points of an insulator detection tool based on solid-state lidar point cloud, comprising the following steps: preprocessing of the solid-state radar point cloud; point cloud clustering; calculating the pose T of the signboard f ; extracting the handle point cloud P h , and calculating the pose T of the grasping point h . The present invention uses technical means of point cloud data processing and feature extraction to calculate the pose information of the insulator detection tool in the working space, and solves the problem that the traditional template matching method cannot be used for the identification and positioning of the insulator detection tool when the similarity between the point cloud shape and the template is low.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer vision, and particularly relates to a method for identifying and positioning the grasping points of an insulator detection tool based on solid-state lidar point cloud. Background Technique

[0002] With the continuous development of lidar technology, the technology of identifying and positioning specific target objects from the three-dimensional point cloud data collected by lidar has been widely applied in the fields of robot perception, autonomous driving, etc. Common methods for identifying objects from three-dimensional point clouds include template matching based on feature description and the iterative closest point algorithm. Both methods rely more on the similarity between the template and the target object in the actual point cloud to be detected. However, due to the influence of factors such as the external environment, scanning perspective, and limitations of the scanning device itself, the complete surface point cloud of the insulator detection tool cannot be obtained in the point cloud to be detected, and the missing part is not fixed, so a unified template cannot be used for matching. Summary of the Invention

[0003] Purpose of the Invention: To solve the technical problems proposed in the background technique, the present invention discloses a method for identifying and positioning the grasping points of an insulator detection tool based on solid-state lidar point cloud.

[0004] Technical Solution: The method for identifying and positioning the grasping points of an insulator detection tool based on solid-state lidar point cloud disclosed by the present invention includes the following steps:

[0005] S1. Preprocessing of solid-state radar point cloud;

[0006] S2. Point cloud clustering

[0007] S2-1. Calculate the normal vectors of all points in the point cloud P o , N o ={n i │i = 1, 2 ⋯ s}, where s is the size of the point cloud P o ;

[0008] S2-2. Through normal vector filtering, retain the solid surface point cloud facing the radar, that is, the X-component n i (n ix， n iy， n iz ) of < - 0.8; ix

[0009] S2-3. Obtain the set Q o of solid surface point cloud clusters through Euclidean clustering;

[0010] S3. Calculate the pose T of the calibration board f

[0011] S3-1. According to the size of the signboard, select the point cloud clusters q that meet the conditions from the set Q of point cloud clusters on the entity surface o ; f ;

[0012] S3-2. Extract the boundary point cloud P on the entity surface by the threshold of the included angle of all point normal vectors of the signboard point cloud. When the included angle between the point normal vector n b and the normal vectors {n1⋯n i} of k b adjacent points is greater than 150°, this point is considered a boundary point, where k k is an adjustable parameter of the boundary extraction algorithm; b

[0013] S3-3. Using the principle of Hough transform, with the centroid of the signboard point cloud as the center, fit the upper boundary normal vector of the signboard boundary point cloud, denoted as the Z vector Z in the signboard coordinate system f . The Hough transform accumulator uses the reciprocal of the standard deviation of the distances d l = {d i │ i = 1, 2 ⋯ n} from the qualified points to the line to be fitted. The line to be fitted is represented by the distance r b from the center to the boundary and the rotation angle θ b of the normal vector around the X axis;

[0014] S3-4. Fit the plane where the signboard boundary point cloud is located by the random sampling method. The plane normal vector n fp is denoted as the X vector X in the signboard coordinate system f ;

[0015] S3-5. Calculate the centroid of the signboard point cloud q f , denoted as the origin O of the signboard coordinate system f . Obtain the Y vector Y in the signboard coordinate system f . Y f = Z f × X f . Finally, obtain the pose T f (O f , X f , Y f , Z f );

[0016] S4. Extract the handle point cloud P h , and calculate the grasping pose T h .

[0017] Among them, in S1, specifically, the point cloud outside the recognition target range is removed by point cloud cropping; voxel filtering is used for downsampling to remove duplicate points within the voxel.

[0018] Further, the screening condition in S3-1 is that the length and width ql of the bounding box of the point cloud clusterf , qw f ∈ [εL, (1 + ε)L] (ε ∈ (0, 1)). At the same time, for the point cloud direction vector (qv major , qv middle , qv minor ), |qv minor .x| > 0.9.

[0019] Furthermore, S4 includes:

[0020] S4-1. Based on the signboard size, installation position, and handle size, with the signboard pose T f as the basis, cut the initial point cloud P o to extract the point cloud of the handle part P h ;

[0021] S4-2. Use the erosion algorithm to remove non-solid noise points and extract the solid point cloud of the handle area P h ={p i |i = 1, 2 ⋯ j}. Traverse the points in the point cloud P h . If there are k i adjacent points within the range of the radius r e centered at p e , then retain this point in the point cloud P h . Here, r e and k e are adjustable parameters of the erosion algorithm;

[0022] S4-3. Cut the point cloud that meets the handle thickness Z h from the highest point in the Z direction of the point cloud P hb to obtain the point cloud of the grasping part P hb ;

[0023] S4-4. Calculate the centroid of the point cloud P hb and record it as the grasping point position O h . Record the main eigenvector V1 of the point cloud P hb as the Y vector Y h of the grasping point coordinate system. Take the X direction of the signboard coordinate system as the X vector X h of the grasping point coordinate system. Finally, obtain the grasping point pose T h (O h , X h , Y h , Z h ), Z h = X h × Y h .

[0024] Beneficial effects: Compared with the prior art, the present invention uses technical means of point cloud data processing and feature extraction to calculate the pose information of the insulator detection tool in the working space, solves the problem that the traditional template matching method cannot be used for the identification and positioning of the insulator detection tool when the similarity between the point cloud shape and the template is low, and provides an important perception scheme for the subsequent operation. Description of the drawings

[0025] Figure 1 It is the implementation flowchart of the present invention;

[0026] Figure 2 It is the point cloud of the detection tool of the present invention and its normal;

[0027] Figure 3 It is the schematic diagram of the point cloud of the sign board, the boundary point cloud and the coordinate system of the sign board of the present invention;

[0028] Figure 4 It is the schematic diagram of the corrosion-filtered point cloud of the present invention;

[0029] Figure 5 It is the schematic diagram of the cut point cloud of the handle and the coordinate system of the grasping point of the present invention. Detailed implementation manners

[0030] The present invention will be further described below in conjunction with the drawings and embodiments.

[0031] The input point cloud used in the present invention is the synthesized point cloud obtained after integrating the solid-state lidar for 3 s. The radar scanning direction is consistent with the direction of the insulator string where the detection tool to be detected is located, and the radar coordinate system is the reference coordinate system.

[0032] As Figure 1 shown, the method for identifying and positioning the grasping point of the insulator detection tool based on the solid-state lidar point cloud includes the following steps:

[0033] S1. Preprocessing of the solid-state radar point cloud.

[0034] Specifically, the point cloud outside the recognition target range is removed by point cloud cropping;

[0035] Voxel filtering for downsampling to remove duplicate points within the voxel.

[0036] S2. Point cloud clustering

[0037] Specifically, it includes:

[0038] S2-1. Calculate the normal of all points in the point cloud P o as shown in Figure 2 N o ={n i │i = 1, 2 ⋯ s}, where s is the size of the point cloud P o ;

[0039] S2-2. Retain the point cloud on the entity surface facing the radar through normal filtering, i.e., the normal vector n i (n ix， n iy， n iz ) The X-component n ix < -0.8;

[0040] S2-3. Obtain the set Q of entity surface point cloud clusters through Euclidean clustering o .

[0041] S3. Calculate the pose T of the signboard f

[0042] Specifically, it includes:

[0043] S3-1. According to the signboard size, screen the qualified point cloud clusters q o from the set Q of entity surface point cloud clusters f ; The screening condition is that the length ql f , width qw f of the bounding box of the point cloud cluster ∈ [εL, (1 + ε)L] (ε ∈ (0, 1)). At the same time, |qv major , qv middle , qv minor ) in the point cloud direction vector > 0.9; minor .x|

[0044] S3-2. Extract the entity surface boundary point cloud P through the threshold of the included angle between the normals of all points in the signboard point cloud b , when the point normal vector n i and the normal vectors {n1⋯n b} of k k adjacent points have an included angle greater than 150°, then this point is considered a boundary point, where k b is an adjustable parameter of the boundary extraction algorithm;

[0045] S3-3. Using the principle of Hough transform, with the centroid of the signboard point cloud as the center, fit the upper boundary normal vector of the signboard boundary point cloud, denoted as the Z vector Z of the signboard coordinate system f , and the Hough transform accumulator uses the reciprocal of the standard deviation of the distances d l = {d i │ i = 1, 2⋯n} from the qualified points to the line to be fitted, where the line to be fitted is represented by the distance r b from the center to the boundary and the rotation angle θ b of the normal around the X-axis;

[0046] S3-4. Fit the plane where the signboard boundary point cloud is located through the random sampling method, and the plane normal vector n fp is denoted as the X direction X of the signboard coordinate systemf ;

[0047] S3-5. As shown in Figure 3 , calculate the centroid of the signboard point cloud q f , denoted as the origin O of the signboard coordinate system f , and obtain the Y vector Y of the signboard coordinate system f , Y f = Z f × X f , and finally obtain the signboard pose T f (O f , X f , Y f , Z f );

[0048] S4. Extract the handle point cloud P h , and calculate the grasping point pose T h .

[0049] Specifically, it includes:

[0050] S4-1. Based on the signboard size, installation position, and handle size, and taking the signboard pose T f as the basis, crop the initial point cloud P o to extract the handle part point cloud P h ;

[0051] S4-2. Use the erosion algorithm to remove non-entity noise points and extract the entity point cloud P of the handle area h = {p i | i = 1, 2 ⋯ j}, as shown in Figure 4 , traverse the points in the point cloud P h . If there are k i neighboring points within the range of the radius r e centered on p e , then retain this point in the point cloud P h , where r e and k e are adjustable parameters of the erosion algorithm;

[0052] S4-3. Intercept the point cloud that meets the handle thickness Z h from the highest point in the Z direction of the point cloud P hb to obtain the grasping part point cloud P hb ;

[0053] S4-4. As shown in Figure 5 , calculate the centroid of the point cloud P hb , denoted as the grasping point position O h , and the main eigenvector V1 of the point cloud P hb is denoted as the Y vector Y of the grasping point coordinate system h, take the X - direction of the sign - board coordinate system as the X - vector X of the grasping - point coordinate system h , and finally obtain the grasping position and orientation T h (O h , X h , Y h , Z h ), Z h = X h ×Y h .

Claims

1. A method for identifying and positioning the grasping points of an insulator detection tool based on the point cloud of a solid-state lidar, characterized in that, It includes the following steps: S1. Solid-state radar point cloud preprocessing; S2. Point cloud clustering S2-1. Calculate the normal vectors of all the points in the point cloud P o , N o = {n i │ i = 1, 2 ⋯ s}, where s is the size of the point cloud P o ; S2-2. Through normal filtering, retain the point cloud of the entity surface facing the radar, i.e., the normal vector n i (n ix， n iy， n iz ) The X-component n ix < -0.8; S2-3. Obtain the set Q of entity surface point cloud clusters through Euclidean clustering o ; S3. Calculate the pose T of the signboard f S3-1. According to the size of the sign board, select the point cloud clusters q that meet the conditions from the set Q of point cloud clusters on the entity surface o ; f ; S3-2. Extract the solid surface boundary point cloud P by the normal angle threshold of all points in the signboard point cloud b , when the point normal vector n i and the normal vectors {n1⋯n b} of k k adjacent points are all greater than 150°, then this point is considered as a boundary point, where k b is an adjustable parameter of the boundary extraction algorithm; S3-3. Using the principle of Hough transform, with the centroid of the signboard point cloud as the center, fit the upper boundary normal vector of the signboard boundary point cloud, denoted as the Z vector Z in the signboard coordinate system f , and the Hough transform accumulator uses the distance d from the qualified points to the line to be fitted l ={d i │i = 1, 2 ⋯ n}, the reciprocal of the standard deviation, where the line to be fitted is represented by the distance r from the center to the boundary b and the rotation angle θ of the normal vector around the X-axis b ; S3-4. Fit the plane where the boundary point cloud of the signboard is located by the random sampling method, and the plane normal vector n fp is denoted as X in the X direction of the signboard coordinate system f ; S3-5. Calculate the centroid of the signboard point cloud q f , denoted as the origin O of the signboard coordinate system f , and obtain the Y vector Y of the signboard coordinate system f , Y f = Z f × X f , and finally obtain the signboard pose T f (O f , X f , Y f , Z f ); S4. Extract the handle point cloud P h , and calculate the grasping position and orientation T h .

2. The method for identifying and positioning the grasping points of the insulator detection tool based on the solid-state lidar point cloud according to claim 1, wherein: In S1, specifically, the point cloud outside the recognition target range is removed by point cloud clipping; voxel filtering for downsampling is performed to remove duplicate points within the voxel.

3. The method for identifying and positioning the grasping points of the insulator detection tool based on the solid-state lidar point cloud according to claim 1, wherein, The screening condition in S3-1 is that the length ql and width qw of the bounding box of the point cloud cluster f , qw f ∈[εL, (1 + ε)L], ε ∈ (0, 1). At the same time, for the point cloud direction vector (qv major , qv middle , qv minor ), |qv minor .x| > 0.

9.

4. The method for identifying and positioning the grasping points of the insulator detection tool based on the solid-state lidar point cloud according to claim 1, characterized in that S4 It includes: S4-1. Based on the signboard size, installation position, and handle size, and with the signboard pose T f as the basis, cut the initial point cloud P o Extract the point cloud P of the handle part h ; S4-2. Use the erosion algorithm to remove non-entity noise and extract the entity point cloud P of the handle area h ={p i |i = 1, 2 ⋯ j}, if there are k i neighboring points within the range of the radius r e centered at p e , then retain this point in the point cloud P h . Among them, r e and k e are adjustable parameters of the erosion algorithm; S4-3. Intercept the point cloud that meets the handle thickness Z downward from the highest point in the Z direction of the point cloud P h to obtain the grasping part point cloud P hb ; hb ; S4-4. Calculate the point cloud P hb The centroid is denoted as the position O of the grasping point h , for the point cloud P hb The main eigenvector V1 is denoted as the Y vector Y of the grasping point coordinate system h , take the X direction of the fiducial board coordinate system as the X vector X of the grasping point coordinate system h , and finally obtain the grasping point pose T h (O h , X h , Y h , Z h ), where Z h = X h × Y h .

Images