CN113627548A

CN113627548A - Planar workpiece template matching method, device, medium and computer equipment

Info

Publication number: CN113627548A
Application number: CN202110940900.4A
Authority: CN
Inventors: 田希文; 高磊
Original assignee: Seizet Technology Shenzhen Co Ltd
Current assignee: Seizet Technology Shenzhen Co Ltd
Priority date: 2021-08-17
Filing date: 2021-08-17
Publication date: 2021-11-09

Abstract

The invention provides a plane workpiece template matching method, which comprises the following steps: s1 obtaining actual workpiece contour point cloud C_sAnd workpiece template contour point cloud C_t(ii) a S2, respectively acquiring the actual workpiece contour point cloud C_sAnd the workpiece template contour point cloud C_tCharacteristic normal vector N of_i＝(n_ix，n_iy，n_iz) Removing the abnormal points based on the characteristic normal vector pair to obtain the actual workpiece contour point cloud C_sAnd workpiece template contour point cloud C_tSampling to obtain point cloud C_tsAnd point cloud C_ss(ii) a S3 aligning the point cloud C_tsAnd the point cloud C_ssPerforming coarse matching positioningObtaining a plurality of coarse matching transformation matrices T_i(i ═ 0.. times, K), and based on each of the transformation matrices T_iFor the point cloud C_tPerforming rotational translation to obtain K point clouds C after transformation_ti(i ═ 0.., K); obtaining a plurality of template contour point clouds S_c(ii) a S4 matching multiple said template contour point clouds S_cAnd the actual workpiece contour point cloud C_sTransformation matrix T for performing fine matching positioning and outputting optimal matching result_bAnd the matching precision and the calculation efficiency can be effectively improved.

Description

Planar workpiece template matching method, device, medium and computer equipment

Technical Field

The invention belongs to the field of image recognition, and particularly relates to a method and a device for matching a planar workpiece template, a storage medium and computer equipment.

Background

In the fields of computer vision and robot perception, 3D point cloud template matching is a common means of object identification and localization. In scenes such as robot carrying, cutting, polishing and the like, the position and the posture of an actual object can be accurately obtained through 3D point cloud template matching, and a calculation basis is provided for robot track generation of subsequent operation tasks. At present, 3D point cloud template matching is generally divided into two stages of rough matching and fine matching, wherein a PPF method commonly used for rough matching is used for carrying out global positioning on an object, and an ICP method commonly used for fine matching is used for further optimizing the result of the rough matching. The traditional PPF method has high degree of dependence on the normal vector characteristics of point cloud, and for a plane object, because the normal vector difference is low, error matching is easy to cause, and the global positioning effect is difficult to play; in addition, because the data volume of the original point cloud is large, the direct adoption of the PPF + ICP method can lead to long calculation time. The accuracy is insufficient and the calculation efficiency is low, so that the traditional PPF + ICP method is difficult to be applied to template matching of a planar object.

Disclosure of Invention

The invention provides a planar object template matching method, which is used for solving the problems.

The invention discloses a plane workpiece template matching method, which comprises the following steps:

s1 obtaining actual workpiece contour point cloud C_sAnd workpiece template contour point cloud C_t；

S2, respectively acquiring the actual workpiece contour point cloud C_sAnd the workpiece template contour point cloud C_tCharacteristic normal vector N of_i＝(n_ix，n_iy，n_iz) Removing the abnormal points based on the characteristic normal vector pair to obtain the actual workpiece contour point cloud C_sAnd workpiece template contour point cloud C_tSampling to obtain point cloud C_tsAnd point cloud C_ss；

S3 aligning the point cloud C_tsAnd the point cloud C_ssCarrying out coarse matching positioning to obtain a plurality of coarse matching transformation matrixes T_i(i ═ 0.., K), and based onEach of the transformation matrices T_iFor the point cloud C_tPerforming rotational translation to obtain K point clouds C after transformation_ti(i ═ 0.., K); k point clouds C_tiOptionally rotating around x-axis and z-axis, or adding K point clouds C_tiTurning around the y axis and rotating around the z axis according to conditions to obtain a plurality of template contour point clouds S_c；

S4 matching multiple said template contour point clouds S_cAnd the actual workpiece contour point cloud C_sCarrying out fine matching positioning, and outputting a transformation matrix of an optimal matching result to be recorded as T_b。

The invention also provides a plane workpiece template matching device, which comprises:

a point cloud obtaining module for obtaining the actual workpiece contour point cloud C_sAnd workpiece template contour point cloud C_t；

A sampling module for respectively acquiring the actual workpiece contour point clouds C_sAnd the workpiece template contour point cloud C_tCharacteristic normal vector N of_i＝(n_ix，n_iy，n_iz) Removing the abnormal points based on the characteristic normal vector pair to obtain the actual workpiece contour point cloud C_sAnd workpiece template contour point cloud C_tSampling to obtain point cloud C_tsAnd point cloud C_ss；

A rough matching module for matching the point cloud C_tsAnd the point cloud C_ssCarrying out coarse matching positioning to obtain a plurality of coarse matching transformation matrixes T_i(i ═ 0.. times, K), and based on each of the transformation matrices T_iFor the point cloud C_tPerforming rotational translation to obtain K point clouds C after transformation_ti(i ═ 0.., K); k point clouds C_tiOptionally rotating around x-axis and z-axis, or adding K point clouds C_tiTurning around the y axis and rotating around the z axis according to conditions to obtain a plurality of template contour point clouds S_c；

A fine matching module for matching a plurality of the template contour point clouds S_cAnd the actual workpiece contour point cloud C_sFor fine matching and positioning and outputting optimum matching resultThe transformation matrix is denoted as T_b。

The invention also provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of any of the methods described above when executing the computer program.

The invention also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, carries out the steps of any of the methods described above.

The invention discloses a plane workpiece template matching method, a device, a storage medium and computer equipment, which are used for matching an original actual workpiece outline point cloud C based on a characteristic normal vector_sAnd workpiece template contour point cloud C_CSampling processing is carried out, the data volume is reduced, and key characteristics of the point cloud are reserved, so that the matching precision and the calculation efficiency are improved. In addition, in the plane workpiece template matching method, the plane workpiece template matching device, the plane workpiece template matching storage medium and the computer equipment, the characteristics of the plane object contour point cloud can be effectively described through self-defining the characteristic normal vector, so that the matching precision and the calculation efficiency are further improved. Meanwhile, the actual workpiece contour point cloud C is obtained_sIn the method, the device, the storage medium and the computer equipment for matching the planar workpiece template, the original discrete point cloud is converted into the four-corner grid with the topological structure, the points at the edge of the whole grid are found by utilizing the topological relation of the four-corner grid, namely the outline of the object, the method does not depend on the difference of normal vectors of the point cloud, only needs the special topological structure of the four-corner grid, and is high in outline extraction efficiency and precision.

Drawings

FIG. 1 is a flowchart illustrating an embodiment of a template matching method according to the present invention;

FIG. 2 is a schematic diagram of a line offset calculation structure in the line of the data structure;

FIG. 3 is a schematic diagram of the structure of arc offset calculation in the data structure arc;

FIG. 4 is a schematic diagram of a circular offset calculation structure in the data structure circle;

FIG. 5 is a workflow diagram of an embodiment of generating an actual workpiece contour point cloud;

FIG. 6 is a flowchart of the operation of another embodiment of the point cloud of the actual workpiece profile;

FIG. 7 is a schematic structural diagram illustrating an embodiment of a template matching apparatus according to the present invention;

fig. 8 is a schematic structural diagram of a computer device according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of the invention in any way.

Like reference numerals refer to like elements throughout the specification. The expression "and/or" includes any and all combinations of one or more of the associated listed items. In the drawings, the thickness, size, and shape of an object have been slightly exaggerated for convenience of explanation. The figures are purely diagrammatic and not drawn to scale.

It will be further understood that the terms "comprises," "comprising," "includes," "including," "has," "includes" and/or "including," when used in this specification, specify the presence of stated features, steps, integers, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, steps, integers, operations, elements, components, and/or groups thereof.

The terms "substantially", "about" and the like as used in the specification are used as terms of approximation and not as terms of degree, and are intended to account for inherent deviations in measured or calculated values that would be recognized by one of ordinary skill in the art.

Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

Example one

As shown in fig. 1, the present invention discloses a template matching method, comprising the following steps:

S3 aligning the point cloud C_tsAnd the point cloud C_ssCarrying out coarse matching positioning to obtain a plurality of coarse matching transformation matrixes T_i(i ═ 0.. times, K), and based on each of the transformation matrices T_iFor the point cloud C_tPerforming rotational translation to obtain K point clouds C after transformation_ti(i ═ 0.., K), K point clouds C are added to the K point clouds_tiOptionally rotating around x-axis and z-axis, or adding K point clouds C_tiTurning around the y axis and rotating around the z axis according to conditions to obtain a plurality of template contour point clouds S_c；

S4 matching multiple said template contour point clouds S_cAnd the actual workpiece contour point cloud C_sCarrying out fine matching positioning, and outputting a transformation matrix of an optimal matching result to be recorded as T_bAnd the matching result is used as the matching result.

The invention discloses a template matching method, which is used for matching original actual workpiece contour point cloud C based on characteristic normal vector_sAnd workpiece template contour point cloud C_tSampling processing is carried out, the data volume is reduced, and key characteristics of the point cloud are reserved, so that the matching precision and the calculation efficiency are improved.

As a preferenceAlternatively, in step S1, the workpiece template contour point cloud C may be obtained based on the following steps_t。

S11A, analyzing the 2D drawing of the workpiece template, and converting the straight line entity, the arc entity and the circle entity of the 2D drawing into a line data structure, an arc data structure and a circle data structure according to preset rules respectively to obtain a plurality of template reference lines S_i(i＝1，2，..，N_s)，N_sRepresenting the total number of the reference lines;

S12A resolving all the line data structure, arc data structure and circle data structure according to the given resolution l₀Respectively generating corresponding reference line point clouds;

S13A merges all the template point clouds to obtain the workpiece template point cloud.

Further, in step S11A, a 2D drawing of a workpiece template is obtained, preset rules of a straight line, an arc and a circle entity in the workpiece template are respectively analyzed into data structures of a line, an arc and a circle according to line attributes of the 2D drawing file, wherein the line, the arc and the circle after analysis are groove reference lines, and the groove reference lines are collected to generate a template reference line S_i(i＝1，2，..，N_s)，N_sThe total number of reference lines is indicated.

In this embodiment, for a specific workpiece template, the 2D drawing file may be in a DXF or DWG format, and the open source library dxfrw may be used to identify line attributes of the 2D drawing file. After the line attributes are obtained, the identified straight line, circular arc and circular solid are respectively analyzed into data structures of line, arc and circle according to preset rules, and then the groove reference lines are formed.

The line member variables comprise the 3D coordinates of the straight line starting point and the 3D coordinates comprising the straight line terminal point, and the line member functions comprise one or more of a line deviation calculation function, a straight line middle point calculation function, a straight line near point calculation function, a straight line far point calculation function, a straight line near point replacement function, a straight line far point replacement function and a straight line starting point terminal point interchange function;

the membership variable of the arc comprises one or more combinations of an arc starting point 3D coordinate, an arc end point 3D coordinate, an arc center 3D coordinate, an arc radius and an excellent arc, and the membership function of the arc comprises one or more combinations of an arc offset calculation function, an arc middle point calculation function, an arc near point calculation function, an arc far point calculation function, an arc near point replacement function, an arc far point replacement function, an arc starting point end point interchange function and the like;

the member variables of the circle comprise a circle entity center 3D coordinate and a circle entity radius, and the member functions of the circle comprise a circle offset calculation function.

Data Structure line

In this embodiment, the member variables of the line in the data structure include 3D coordinates of a start point and an end point, and the member functions of the line include an offset line calculation function, a middle point calculation function, a near point calculation function, a far point calculation function, a near point replacement function, a far point replacement function, and a start point and end point replacement function.

(a) For a straight line entity in a 2D drawing, the starting point coordinate P of a line_s＝(x_s，y_s，z_s)^TReference point coordinate (x) corresponding to straight line entity_b，y_b) Wherein x is_s＝x_b，y_s＝y_b，z_sLine end point coordinate P of 0_e＝(x_e，y_e，z_e)^TEnd point coordinate (x) of corresponding straight line entity₂，y₂) Wherein x is_e＝x₂，y_s＝y₂，z_s＝0。

(b) The line offset calculation function of the line comprises 2 inputs and 1 output, wherein the input 1 is a designated offset, the input 2 is an offset direction, and the output is the offset line. As shown in fig. 2, the start point and the end point of the line have the same coordinate system, with the x-axis direction pointing along the start point to the end point, the z-axis direction pointing down perpendicular to the workpiece surface, and the x-y-z axes constituting the right-hand coordinate system. After the offset dy is given, if the offset direction is the regular starting point and the regular end point, the offset dy is towards the positive direction of the y axis; if the offset direction is negative, the start point and the end point are offset by dy in the negative y-axis direction.

(c) line midpoint calculation function for confirming line middleThe point, the middle point of the line, is the middle point of the straight line, the 3D coordinate (x) of the middle point_m，y_m，z_m) 3D coordinates with start and end points

x_m＝(x_s+x_e)/2，y_m＝(y_s+y_e)/2，z_m＝(z_s+z_e)/2

(d) The near point calculation function and the far point calculation function of the line are used for determining a near point and a far point of any specified point P, wherein if the distance from the starting point of the line to the specified point P is greater than or equal to the distance from the end point to the specified point P, the starting point is the near point and the end point is the far point for the specified point P; conversely, if the distance from the starting point to the point P is less than the distance from the end point to the point P, the starting point is a far point and the end point is a near point. Wherein, the distance from the designated point P to the starting point and the end point can be respectively calculated according to the Euclidean distance.

(e) A near point replacing function of line, which is used for replacing the near point with the point P aiming at the specified point P; the far point replacement function is used for replacing the far point with a point P; the start point and end point interchange function is used for interchanging the start point and the end point.

(II) data Structure arc

The member variables of the data structure arc comprise 3D coordinates of a starting point, an end point and a circle center, the member functions of the data structure arc comprise a radius and good and bad arc attributes, and function functions of offset circular arc, middle point, near point, far point, near point replacement, far point replacement and starting point and end point interchange are calculated.

(a) For the arc entity in the 2D drawing, the starting point coordinate P of arc_s＝(x_s，y_s，z_s)^TStarting angle theta corresponding to arc entity_sEnd point coordinate P of arc_e＝(x_e，y_e，z_e)^TEnd angle theta corresponding to arc entity_eCenter of circle coordinate P of arc_c＝(x_c，y_c，z_c)^TReference point coordinate (x) corresponding to arc entity_b，y_b) Radius of arc r_aRadius r of corresponding arc entity_bGood and bad arc attribute N of arc_aCorresponding to the solid termination of the arcAngle theta_eFrom the starting angle theta_sThe relationship is as follows:

x_c＝x_b，y_c＝y_b，z_c＝0，r_a＝r_b；

x_s＝x_c+r_a cosθ_s，y_s＝y_x+r_a sinθ_s，z_s＝z_c；

x_e＝x_c+r_a cosθ_e，y_e＝y_c+r_a sinθ_e，z_e＝z_c；

(b) the circular arc offset calculation function of arc is used for confirming the offset arc, and comprises 2 inputs and 1 output, wherein the input 1 is the designated offset, the input 2 is the offset direction, and the output is the offset arc. As shown in fig. 3, after the offset dr is given, if the offset direction is a regular starting point and an end point, the offset dr is shifted in a direction in which the radius increases; if the offset direction is negative, the start point and the end point are offset dr in a direction in which the radius decreases.

(c) The midpoint calculation function of arc is used to determine the midpoint of arc, in this embodiment, the midpoint of arc refers to the midpoint of arc, and the 3D coordinate P of the midpoint_m＝(x_m，y_m，z_m)^TWith the starting points (xs, ys, zs) and) x_e，y_e，z_e) The endpoint satisfies the following relationship:

x_cs＝x_c-x_s，y_cs＝y_c-y_s，z_cs＝z_c-z_s；

x_ce＝x_c-x_e，y_ce＝y_c-y_e，z_ce＝z_c-z_e；

θ₁＝atan2(y_cs，x_cs)，θ₂＝atan2(y_ce，x_ce)，Δθ＝θ₂-θ₁；

P_m＝P_c+R_z(θ_m)(P_c-P_s)；

wherein R is_z(θ_m) For rotation of theta about z-axis_mThe rotation matrix of (2).

(d) The arc near point calculation function and the far point calculation function are used for determining the near point and the far point of the designated point P, wherein if the distance from the starting point to the point P is greater than or equal to the distance from the end point to the point P, the starting point is the near point of the designated point P, and the end point is the far point of the designated point P; conversely, if the distance from the starting point to the point P is less than the distance from the end point to the point P, the starting point is the far point of the point P, and the end point is the near point of the point P, wherein the distance from the starting point to the point P can be calculated by the Euclidean distance

(e) The near point replacement function of arc is used to replace the near point with point P; the far point replacing function is used for replacing the far point with a point P; the starting point and end point interchange function is used for interchanging the starting point and the end point according to actual requirements under the condition that the positions of the starting point and the end point are interchanged after the analysis is finished.

(III) data Structure circle

circle contains the 3D coordinates and radius of the center of the circle and calculates the function of the offset circle.

(a) For a circular solid in a 2D drawing, the circle center coordinate P of a circle_c＝(x_c，y_c，z_c)^TReference point coordinate (x) corresponding to arc entity_b，y_b) Radius r of circle_cRadius r of corresponding round solid_bThe relationship is as follows:

x_c＝x_b，y_c＝y_b，z_c＝0，r_c＝r_b

(b) the circle offset calculation function of circle is used for confirming offset circle, and comprises 2 inputs and 1 output, wherein the input 1 is a designated offset, the input 2 is an offset direction, and the output is offset circle. As shown in fig. 4, after the offset dr is given, if the offset direction is the regular direction in which the whole circle is offset dr toward the direction in which the radius is increased; if the offset direction is negative, the entire circle is offset dr in a direction in which the radius decreases.

Respectively analyzing entities with attributes of straight lines, circular arcs and circles in the 2D drawing file into specific data structures of lines, arc and circle, and then obtaining template groove lines of the workpiece template; on the basis, the flow proceeds to step S12a, where the line, arc and circle of the workpiece template are processed according to the given resolution l₀Respectively generating corresponding reference line point clouds Mi, and combining all the point clouds Mi to obtain a workpiece template point cloud C_t。

In step S12A, for any one line data structure, the corresponding point cloud Mi can be generated according to the following method: starting from 0 every interval l₀Adding points P to a point cloud M_iAnd if l₀If l cannot be divided completely, the point P is divided_eIs also added to the point cloud Mi, where P_i＝P_s+k_id，k_i＝i·l₀(0≤k_iL) is not more than l), wherein the point cloud Mi is a pre-created empty point cloud, k represents the number in the point cloud Mi, d represents the direction of line,

l represents the length of line, and l | | | P_e-P_s||。

In step S12A, for any arc data structure, a corresponding point cloud may be generated according to the following method:

obtaining good and bad arc attributes and rotation angle theta of arc₀And magnitude of Δ θ; reading the central angle corresponding to the arc, wherein if the central angle is not less than 180 degrees, the central angle is an major arc, and if the central angle is less than 180 degrees, the central angle is a minor arc;

if the arc is judged to be the major arc and the delta theta is more than or equal to 0, starting from 0 and every interval theta₀Adding a point P to the point cloud M_i，

P_i＝P_c+R_z(θ_i)，θ_i＝i·θ₀(0≤θ_i≤2π-Δθ)；

If the arc is judged to be the major arcAnd Δ θ<0, then every interval-theta starting from 0₀Adding a point P to the point cloud M_i，

P_i＝P_c+R_z(θ_i)，θ_i＝-i·θ₀(2π+Δθ≤θ_i≤0)；

If the arc is judged to be a minor arc and the delta theta is more than or equal to 0, starting from 0 every interval theta₀Adding a point P to the point cloud M_i，

P_i＝P_c+R_z(θ_i)，θ_i＝i·θ₀(0≤θ_i≤Δθ)；

If the arc is judged to be a minor arc and delta theta<0, then every interval-theta starting from 0₀Adding point P to the point cloud Mi_i，

P_i＝P_c+R_z(θ_i)，θ_i＝-i·θ₀(Δθ≤θ_i≤0)；

If theta₀If Δ θ cannot be divided completely, the point P is divided_eAlso to the point cloud M.

Wherein the point cloud Mi is a pre-created empty point cloud, and the rotation angle interval theta₀＝l₀/r_aΔ θ is the vector CP_sAnd vector CP_eAngle of (a) vector CP_s＝P_s-P_c，CP_e＝P_e-P_cAnd the value range of delta theta is (-pi, pi).

In step S12A, for any circle data structure, a corresponding point cloud can be generated according to the following method: obtaining a rotation angle interval theta₀＝l₀/r_cStarting from 0 every interval θ₀Adding point P to the point cloud Mi_iWherein the point cloud Mi is a pre-created empty point cloud

P_i＝P_c+R_z(θ_i)，θ_i＝i·θ₀(0≤θ_i≤2π)。

Merging point clouds generated by all lines, arc and circle, wherein the obtained point cloud is the workpiece template point cloud C_t

In addition, as a preferred embodiment, the element template is obtainedPoint cloud C_tAnd then also comprises: obtaining the rotational inertia of the template point cloud, obtaining a pose matrix of the template point cloud, and establishing a template coordinate system S according to the pose matrix_mTranslating and rotating the template point cloud to a coordinate system S_m。

As a preferred scheme, as shown in fig. 5, in step S1, the actual workpiece contour point cloud C can be obtained based on the following method_s：

S11B obtaining actual workpiece point cloud C₁Based on a preset rule, the actual workpiece point cloud C is processed₁Conversion into a tetragonal grid SM₁And acquire the SM₁Normal vector of each vertex in the vector;

S12B segmenting the SM based on the normal vector and a preset constraint condition₁Removing the SM based on the segmentation result₁Generating a four-corner grid SM after removing the side point cloud₂；

S13B searching and acquiring the SM₂All the edge points of (a), said edge points constituting a point cloud C₃Namely the actual contour point cloud of the workpiece to be extracted.

Further, in step S11B, an original ordered point cloud of the object to be detected is obtained based on the 3D camera, where the point cloud data is recorded in the form of points for the object to be detected, and each point includes three-dimensional coordinates including color information and a normal vector. More specifically, the color information is usually obtained by a camera to obtain a color image, and then the color information of the pixel at the corresponding position is assigned to the corresponding point in the point cloud; the ordered point cloud is typically in the camera coordinate system with the normal facing the camera.

The step S11B of converting the point cloud into a four-corner grid based on a preset rule includes the following steps:

s111 for any point P in the ordered point cloud_ijAdding said P_ijTo obtain vertex V, the 3D position coordinates (x, y, z) normal vector, color information_mnEstablishing said P_ijAnd said vertex V_mnThe index relationship is in one-to-one correspondence;

in this step, a vertex set S is established, a row coordinate i and a column coordinate j of the ordered point cloud are traversed, and P is added_ijAnd acquiring vertex V of corresponding 3D coordinate, normal vector and color information_mnThen, each vertex V is put_mnTo the set of vertices S. The vertex set S is a vertex set of a quadrilateral mesh (seizetccolormesh).

S112, acquiring the row and column sequence of each point in the point cloud data;

one frame of scan data is a depth map, belonging to a gray scale map. The grey value of the pixel represents depth information, each pixel point can be converted to a world coordinate system through camera parameters, so that each pixel corresponds to a three-dimensional point, and if the resolution of the depth map is ResX ResY, the ordered point cloud is arranged from the upper left corner to the lower right corner in a line-by-line manner according to the square matrix of the map. Due to the points P in the ordered point cloud_ijThe points are arranged in sequence, in the embodiment, the point cloud data is traversed, and the row coordinate i and the column coordinate j of each point in the ordered point cloud are read to obtain V_ijSorted by rows and sorted by columns, respectively.

S113, any point P in the ordered point cloud_ijSearching P according to the index rule of the row-column ordering_ijAdjacent point P of_i,j+1,P_i+1,j+1,P_i+1,jWherein, the ith row and j column, the ith row and j +1 column, the ith +1 row and j column and the ith +1 row and j +1 column in the point cloud data respectively correspond to P_i,j,P_i,j+1,P_i+1,j+1,P_i+1,j4 points are selected;

because the ordered point cloud is arranged in sequence, the information of adjacent points can be easily found. In this embodiment, as a preferred scheme, any point P is searched for_ijWhen there is a neighboring point, if P_i,j+1、P_i+1,j+1、P_i+1,jIf any one of the three adjacent points can not be found from the ordered point cloud data, skipping V_ijThe next point neighbor search is continued. The row and column coordinates of the ordered point cloud are range-wide, e.g., (i 0, …, Nr), (j 0, …, Nc), if i<0 or i +1>Nr or j<0 or j +1>Nc, judging that the boundary is exceeded. If the row coordinate i +1 or the column coordinate j +1 exceeds the boundary, the point P is skipped_ijAnd continuing to search the next point in the ordered point cloud corresponding to the three adjacent points.

S114, obtaining P according to the index relation_i,j,P_i,j+1,P_i+1,j+1,P_i+1,jFour vertexes V corresponding to each other respectively_mn，V_m,n+1.V_m+1,n，V_m+1,n+1；

Adding four corner patches and putting the 4 vertexes V_mn，V_m,n+1.V_m+1,n，V_m+1,n+1Adding to vertex positions of corresponding quadrilateral patches to generate half-edge quadrilateral meshes, wherein each vertex V_mnStoring corresponding point P in point cloud data_ijThe 3D position coordinates (x, y, z), normal vectors, colors, and pointer information of the halves of which are the starting vertex, each half storing pointers or indices of the starting vertex, the ending vertex, the adjoining surface, the upper half, the lower half, the opposite half;

s115 traversal of the P_ijOr the vertex V_mnObtaining all of said P_ijOr the vertex V_mnCorresponding half four-corner grids to output four-corner grids SM corresponding to workpiece point cloud data₁。

In the shooting process, the side face of the workpiece can be inevitably shot due to the influence of factors such as the structure of the workpiece, the shooting angle and the like, but in actual application, subsequent operation is often required to be performed based on the front point cloud of the workpiece, for example, when the track generation is performed on the workpiece chamfering, the step S11B is used for converting the point cloud of the workpiece into the corresponding four-corner mesh SM₁Then, the process proceeds to step S12B to remove the side point cloud (i.e., the side with less point cloud).

Step S12B is to the SM based on the normal vector and a preset constraint condition₁Dividing, distinguishing different planes of actual workpiece point clouds, removing the side point clouds of the workpiece based on the dividing result, and finally obtaining the four-corner mesh SM after removing the side point clouds₂。

In one of the embodiments, it can be based directly on the tetragonal grid SM₁Performing point cloud plane segmentation to obtain four-corner mesh SM (sparse space) for removing side point cloud₂The method specifically comprises the following steps:

step (A): traversal SM₁All points in (1) { V }₁，V₂，...V_nN denotes SM₁Number of midpoints, decision point V_mIf it has already been processed, if it is point V_mCreating an empty cluster container without processing_iAnd will V_iIs added to D_iAt the same time, point V_mMarking as processed; if point V_mTo be processed, point V is skipped_m；

For addition to D_iPoint V of_mSimultaneously performing step (a) comprising:

at a predetermined distance d₀Within range, search point V_mNeighbor point of (V)_m1，V_m2，...V_mnN denotes a point V_mAnd judging V_mIs a neighbor point V_mnWhether a preset constraint condition is met or not, wherein the preset constraint condition is a neighbor point V_mnUntreated and V_mAnd V_mnThe included angle of the normal vector is not more than a preset angle threshold value theta₀(ii) a Wherein, if V_mIs a neighbor point V_mnIf the preset constraint condition is met, the point V is set_mnIs also added to the vessel D_iAnd a point V is formed_mnAlso marked as processed; if the neighbor point V_mnAfter processing, skipping point V_mnTo V pair_mThe next neighbor point of (a); if point V_mAt a predetermined distance d₀If there is no neighbor point in the range, then skip point V_mTo the container D_iThe other points in (a) perform step (a) up to SM₁Finishing the execution of all the points;

the processed means that the corresponding point has been added to the clustering container D_iIn (1), repeating the above steps until SM₁After all the points are executed, a plurality of clustering containers { D ] can be obtained₁，D₂，...D_kK represents the number of the clustered containers, and each container corresponds to one point cloud; get { D₁，D₂，...D_kThe point cloud with the largest number of points is the four-corner grid SM after the side point cloud is removed₂。

In another embodiment, as shown in FIG. 6, the row point cloud may also be based on the workpiece point cloud C1Plane segmentation, namely obtaining a workpiece point cloud C2 with the side point cloud removed, and converting the workpiece point cloud C2 into a quadrilateral mesh SeizetcColorMesh SM according to a preset rule₂The method specifically comprises the following steps:

s121, according to the corresponding relation, the SM₁Assigning the normal vector of each vertex to each corresponding point of the workpiece point cloud;

s122, segmenting the C1 based on the normal vector and a preset constraint condition, removing the side point cloud of the workpiece based on the segmentation result, and generating a workpiece point cloud C with the side point cloud removed₂；

S123, based on preset rules, the workpiece point cloud C₂Conversion to the quadrangle grid SeizetColorMesh SM₂。

Further, step S122 specifically includes the following steps:

step (B): traverse point cloud C₁All points in { P }₁，P₂，...P_nN represents the point cloud C₁Number of midpoints, decision point P_iIf it has already been processed, if it is point P_iCreating an empty cluster container without processing_iAnd is combined with P_iIs added to D_iAt the same time, point V_iMarking as processed; if point P_iSkipping point P for processed_i；

At the same time, for addition to D_iPoint P of_iPerforming step (b) comprising: at a predetermined distance d₀Search point within range P_iNeighbor point of { P_i1，P_i2，...P_ijJ denotes a point P_iAnd judges P_iNeighbor point P of_ijWhether a preset constraint condition is met or not, wherein the preset constraint condition is a neighbor point P_ijUntreated and P_iAnd P_ijIs not more than an angle threshold theta₀；

Wherein, if P_iNeighbor point P of_ijIf the preset constraint condition is satisfied, the point P is set_ijIs also added to the vessel D_iAnd connecting the point P_ijAlso marked as processed; if the neighbor point P_ijAfter having been processed, then jumpPassing point P_ijTo P_iThe next neighbor point of (a) continues to perform step (b); if point P_iAt a distance d₀If there is no neighbor point in the range, then skipping point P_iTo the container D_iThe other points in (a) are executed until all the points in the point cloud C1 are executed.

Obtaining a number of clustered containers { D₁，D₂，...D_kK represents the number of the clustered containers, and each container corresponds to one point cloud; get { D₁，D₂，...D_kThe point cloud with the largest number of points is the workpiece point cloud with the side point cloud removed, and is marked as C₂。

The processed means that the corresponding point has been added to the clustering container D_iIn the method, the steps are repeated until all the points in the C1 are executed, and a plurality of clustering containers { D are obtained₁，D₂，...D_kK represents the number of the clustered containers, and each container corresponds to one point cloud; get { D₁，D₂，...D_kThe point cloud with the largest number of points is the workpiece point cloud with the side point cloud removed, and is marked as C₂。

Further, in this embodiment, the actual workpiece point cloud C is processed₁Generating a Kd search tree, based on which: at a predetermined distance d₀Within range, search point P_iNeighbor point of { P_i1，P_i2，...P_ijWherein, for different workpieces, an empirically settable distance threshold d is used₀And an angle threshold theta₀And (4) taking values.

S123, based on preset rules, the workpiece point cloud C₂Conversion to the quadrangle grid SeizetColorMesh SM₂. Generating the corresponding four-corner mesh based on the point cloud as described in step S1 may include the steps of:

for C in the point cloud₂Any point P of_ijAdding said P_ijTo obtain vertex V, the 3D position coordinates (x, y, z) normal vector, color information_mnEstablishing said P_ijAnd said vertex V_mnThe index relationship is in one-to-one correspondence;

obtaining the actual workPiece point cloud C₂Sequencing the rows and columns of each point in the sequence;

for any point P in the point cloud_ijSearching P according to the index rule of the row-column ordering_ijAdjacent point P of_i,j+1,P_i+1,j+1,P_i+1,jWherein, the ith row and j column, the ith row and j +1 column, the ith +1 row and j column and the ith +1 row and j +1 column in the point cloud data respectively correspond to P_i,j,P_i,j+1,P_i+1,j+1,P_i+1,j4 points are selected;

obtaining P according to the index relation_i,j,P_i,j+1,P_i+1,j+1,P_i+1,jFour vertexes V corresponding to each other respectively_mn，V_m,n+1.V_m+1,n，V_m+1,n+1；

traverse the P_ijOr the vertex V_mnObtaining all of said P_ijOr the vertex V_mnCorresponding half four-corner grids to output four-corner grids SM corresponding to the point cloud data₂。

Obtaining a four-corner grid SM₂Thereafter, the process proceeds to step S13B, where the SM is searched and acquired₂All the edge points are combined into a point cloud C₃，C₃Namely the actual contour of the workpiece to be extracted.

In this embodiment, the tetragonal grid SM is traversed₂For all vertices of the quadrilateral mesh SM₂Vertex V of_i(i0，...，n₂) If V is_iIf one of the following conditions is satisfied, V is judged_iIs SM₂The edge points of (a):

vertex V_iOut half E of_jIs invalid;

(II) half E_jCorresponding patch Q_kIs invalid;

(III) vertex V_iOut half E of_jIs effective, but E_jIs SM₂Wherein, if E_jIf the condition (b) is satisfied, E is judged_jIs SM₂The edge of (1).

Of course, in the present embodiment, the actual workpiece contour point cloud C is not clear_sAnd workpiece template contour point cloud C_t(ii) a The acquisition may be replaced by other methods, not limited to the method based on the tetragonal grid, such as a contour extraction method based on direct point cloud.

Respectively carrying out actual workpiece contour point cloud C based on step S1_sAnd workpiece template contour point cloud C_tThen, the process goes to step S2 to obtain the actual workpiece contour point clouds C respectively_sAnd the workpiece template contour point cloud C_tCharacteristic normal vector N of_i＝(n_ix，n_iy，n_iz) Removing the abnormal points based on the characteristic normal vector pair to obtain the actual workpiece contour point cloud C_sAnd workpiece template contour point cloud C_tSampling to obtain point cloud C_tsAnd point cloud C_ssThe method comprises the following steps:

s21 point cloud C_tAnd point cloud C_sThe characteristic normal vector of (1);

s22 removing point cloud C_tAnd point cloud C_sA point of medium abnormal;

s23 Point cloud C based on feature normal vector (after removing abnormal points)_tAnd point cloud C_sSampling to obtain point cloud C_tsAnd point cloud C_ss；

In this embodiment, as a preferred scheme, in step S21, the actual workpiece contour point cloud C_sAnd workpiece template contour point cloud C_tThe characteristic normal vector acquisition comprises the following steps: obtaining the geometric center point P of the point cloud C_centerGo through each point P of the point cloud C_iFor each point P_iIts characteristic normal vector

The abnormal points refer to points with a value of NaN (not a No.), and as the result of binary calculation (addition, subtraction, multiplication and division) between NaN and any number is that NaN affects the calculation result if the abnormal points are not removed and participate in subsequent calculation, the point cloud C is removed in step S22 respectively_tAnd point cloud C_sAs a preferred scheme, in this embodiment, when the point cloud C is abnormal, the point cloud C is obtained_tAnd point cloud C_sPoint P in_iWhen any one of the following conditions is met, the point is regarded as an abnormal point, and the abnormal point is removed from the corresponding point cloud;

(a) if point P_iIn the three-dimensional coordinates (x, y, z), x or y or z is NaN (not a No.), then point P is_iIs an abnormal point;

(b) if point P_iCharacteristic normal vector N of_i＝(n_ix，n_iy，n_iz) In, n_ixOr n_iyOr n_izIs NaN, then point P_iIs an abnormal point;

since the number of point clouds is large, in order to find the feature points more quickly and accurately for the subsequent matching step, in this embodiment, step 23 is to match the point cloud C based on the feature normal vector_tAnd point cloud C_sSampling is performed. As a preferred scheme, the point clouds C can be respectively processed based on the following method_tAnd point cloud C_sSampling:

step (C): traverse all points { V ] in the point cloud C₁，V₂，...V_nN represents the number of points in the point cloud C, and a judgment point V_iIf it has already been processed, if it has not, V_iAdding to the point cloud C_oAnd will point V_iMarked as processed, wherein C_oIs a pre-created empty point cloud; if point V_iProcessed, then skipping point V_i；

For addition to C_oPoint V in_iSynchronously performing step (c) comprising: at a predetermined distance d₀Search point within range V_iNeighbor point of (V)_i1，V_i2，...V_imM denotes a point V_iAnd judging V_iIs a neighbor point V_ijWhether a preset constraint condition is met or not, wherein the preset condition is as follows: neighbor point V_ijUntreated and V_iAnd V_ijAre not more than a preset angle threshold theta₀；

If V_iIts neighbor point V_ijAfter processing, skipping point V_ijJudgment of V_iWhether other neighbor points of (1) satisfy the constraint condition;

if V_iIts neighbor point V_ijUnprocessed, decision point V_ijIf the constraint condition is satisfied, if the point V is_iWith its neighbour point V_ijIf the preset constraint condition is met, the point V is set_ijAlso added to the point cloud C_oAnd will point V_ijAlso marked as processed;

if point V_iAt a distance d₀If there is no neighbor point in the range, then skip point V_iPoint-to-point cloud C_oThe other point of (a) performs step (C).

Further, in the step (C), a Kd search tree is generated according to the point cloud C, and the Kd search tree is used at a preset distance d₀Search point within range V_iNeighbor point of (V)_i1，V_i2，...V_im}。

Point-to-point cloud C through step S2_tAnd point cloud C_sSampling is carried out to respectively obtain point clouds C after sampling_tsAnd point cloud C_ssThen, considering that a certain deflection angle exists between the workpiece template point cloud and the actual workpiece point cloud in the placing posture, if the sampled point cloud C is directly used_tsAnd point cloud C_ssThe matching is carried out, the probability of successful matching is very low, so that the scheme shown in the invention is required to firstly obtain a certain amount of point clouds with converted angles through rough matching, and then output the template contour point cloud with the highest matching proportion as the matching result through accurate matching.

As a preferable solution, in this embodiment, step S3 uses a ppf (point Pair features) registration method to register point cloud C_tsAnd point cloud C_ssPerforming rough matchingA bit, comprising:

s31 acquiring workpiece template C before sampling_tMaximum workpiece radius r_max＝‖P_max-P_minII/2 wherein P_minAnd P_maxFor a workpiece template contour point cloud C_tMinimum point P of_minAnd a maximum point P_max；

S32 is processed by the sampled workpiece template contour point cloud C_tsInput point cloud, point cloud C as PPF feature estimator_ssAs target point cloud, 2r_maxAs a distance threshold of the PPF hash map searcher, a PPF registration method is used to obtain a plurality of coarse matching results, namely transformation matrixes; t is_i(i＝0，...，K)；

S33 uses each transformation matrix T_iPoint-to-point cloud C_tPerforming rotational translation to obtain K transformed point clouds C_ti(i＝0，...，K)；

S34 point cloud C_tiTurning around an x axis and rotating around a z axis according to conditions; or point cloud C_tiTurning around the y axis and rotating around the z axis according to conditions to respectively obtain a plurality of template contour point cloud sets S_c。

Since the actual workpiece placement positions and postures are often different, and even the situation completely opposite to the workpiece template may occur, in step S34, the point cloud C is used_tiConditional wrap transform matrix T_iThe x-axis overturn and the z-axis rotation are carried out to obtain a plurality of template contour point cloud sets S_cThe method comprises the following steps:

s341 according to the point cloud C_ti(i ═ 0.... K.) corresponding transformation matrix T_iCalculating its corresponding rotational z-axis z_i：

z_i＝T_i(0，2，3，1)

Wherein, T_i(0, 2, 3, 1) represents a number from T_iStarting to take 3 numbers from the 0 th row and the 2 nd column to form 1 column;

s342, the point cloud Cti is turned around the x axis according to the condition, if the z axis z is rotated_iContrary to the vertically downward direction, point cloud C is pointed_tiRotated by 180 ° about the x-axis, which is the transformation matrix T_iX axis of rotation x_iI.e. by

x_i＝T_i(0，0，3，1)

Wherein, T_i(0, 0, 3, 1) represents a number from T_iThe 0 th row and the 0 th column of the (1) form 1 column by taking 3 numbers;

s343 point cloud C_tiRotation around the z-axis conditional: point cloud C_ti(i ═ 0.., K) around the rotational z-axis z_，Rotation n_zEach time of rotation angle is

To obtain n_zK new point clouds of which n_zFor a predetermined number of rotation steps, and (n)_z+1) K template contour point clouds are stored in the template point cloud set S_cPerforming the following steps;

or, the point cloud C_tiThe method comprises the following steps of turning around the y axis and rotating around the z axis according to conditions to obtain a plurality of template contour point clouds:

s341' from the point cloud C_ti(i ═ 0.... K.) corresponding transformation matrix T_iCalculating its corresponding rotational z-axis z_i：

z_i＝T_i(0，0，1，1)；

Wherein, T_i(0, 2, 3, 1) represents a number from T_iThe 0 th row and the 0 th column of the (1) form 1 column by taking 3 numbers;

s342' Point cloud C_tiOptionally turning around the y-axis if the z-axis is rotated_iContrary to the vertically downward direction, point cloud C is pointed_tiRotated by 180 about the x-axis, the y-axis being the transformation matrix T_iX axis y of rotation of_iI.e. by

y_i＝T_i(0，1，3，1)

Wherein, T_i(0, 0, 3, 1) represents a number from T_iThe number of the row 0 and the column 1 is 3, and 1 column is formed.

S343' Point cloud C_tiConditionally rotating about the z-axis, comprising: point cloud C_ti(i ═ 0.., K) around the rotational z-axis z_，Rotation n_zEach time of rotation angle is

To obtain n_zK new point clouds of which n_zFor a predetermined number of rotation steps, and (n)_z+1) K template contour point clouds are stored in the template point cloud set S_cIn (1).

(n) can be obtained after the processing of steps S341 to S343 or steps S341' to S343_z+1) K template contour point clouds, at this point, proceeding to step S4, and fitting a plurality of template contour point clouds S_cAnd the actual workpiece contour point cloud C_sCarrying out fine matching positioning, and outputting a transformation matrix of an optimal matching result to be recorded as T_bAnd the matching result is used as the matching result.

Preferably, in this embodiment, the set S of ICP pairs is used_cTemplate point cloud and workpiece contour point cloud C in_sPerforming fine matching positioning, including:

s41 traversing set S_cTemplate point cloud C in_jAnd respectively connected with the point cloud C_sPerforming ICP registration, and if the registration is convergent, the matching degree is less than a preset value and the matching proportion is greater than the preset value, determining the point cloud C_jAnd point cloud C_sMatching is successful; obtaining a plurality of matching results which are successfully matched, including a transformation matrix T_jMatching degree and matching proportion; and if the matching result of the matching is not successful, the 3D template matching is considered to be failed.

S42, according to the matching proportion, the matching results in the step S41 are sorted, and the transformation matrix of the output optimal matching result is recorded as T_bThen T is_bAnd matching an optimal result for the 3D plane workpiece template.

Example two

As shown in fig. 7, the present invention also discloses a template matching apparatus 10, comprising:

a point cloud obtaining module 11 for obtaining the actual workpiece contour point cloud C_sAnd workpiece template contour point cloud C_t；

A sampling module 12 for respectively obtaining the actual workpiece contour point clouds C_sAnd the workpiece template contour point cloud C_tCharacteristic normal vector N of_i＝(n_ix，n_iy，n_iz) Removing the abnormal points based on the characteristic normal vector pair to obtain the actual workpiece contour point cloud C_sAnd workpiece template contour point cloud C_tSampling to obtain point cloud C_tsAnd point cloud C_ss；

A rough matching module 13 for matching the point cloud C_tsAnd the point cloud C_ssCarrying out coarse matching positioning to obtain a plurality of coarse matching transformation matrixes T_i(i ═ 0.. times, K), and based on each of the transformation matrices T_iFor the point cloud C_tPerforming rotational translation to obtain K point clouds C after transformation_ti(i ═ 0.., K); k point clouds C_tiOptionally rotating around x-axis and z-axis, or adding K point clouds C_tiTurning around the y axis and rotating around the z axis according to conditions to obtain a plurality of template contour point clouds S_c；

A fine matching module 14 for matching a plurality of said template contour point clouds S_cAnd the actual workpiece contour point cloud C_sCarrying out fine matching positioning, and outputting a transformation matrix of an optimal matching result to be recorded as T_b。

Preferably, the point cloud obtaining module 11 includes an actual workpiece contour point cloud C_sObtaining a sub-module, the actual workpiece contour point cloud C_sThe acquisition submodule is used for acquiring an actual workpiece point cloud C₁Based on a preset rule, the actual workpiece point cloud C is processed₁Conversion into a tetragonal grid SM₁And acquire the SM₁Normal vector of each vertex in the vector; segmenting the SM based on the normal vector and a preset constraint condition₁Removing the SM based on the segmentation result₁Generating a four-corner grid SM after removing the side point cloud₂(ii) a Searching for and acquiring the SM₂All the edge points of (a), said edge points constituting a point cloud C₃Namely the actual contour point cloud of the workpiece to be extracted.

Preferably, the point cloud obtaining module 11 comprises a workpiece template contour point cloud C_tObtaining a sub-module, the workpiece template contour point cloud C_tThe method comprises the steps of obtaining a 2D drawing of a submodule for analyzing a workpiece template, and enabling a straight line entity, an arc entity and a position of the 2D drawing,The round solid is converted into a line data structure, an arc data structure and a circle data structure according to preset rules respectively to obtain a plurality of template reference lines S_i(i＝1，2，..，N_s)，N_sRepresenting the total number of the reference lines; analyzing all the line data structure, the arc data structure and the circle data structure according to a given resolution l₀Respectively dispersing to generate corresponding reference line point clouds; and merging all the reference line point clouds to obtain the workpiece template point cloud.

Preferably, the sampling module 12 includes a feature normal vector obtaining sub-module, and the feature normal vector obtaining sub-module is configured to obtain a geometric center point P of the point cloud C_centerGo through each point P of the point cloud C_iFor each point P_iIts characteristic normal vector

Preferably, the sampling module 12 includes an abnormal point obtaining sub-module, and the abnormal point obtaining sub-module is used for obtaining the point cloud C_tAnd point cloud C_sPoint P in_iA point which is considered abnormal when any one of the following conditions is satisfied;

(a) if point P_iIn the three-dimensional coordinates (x, y, z), x or y or z is NaN, the point P_iIs an abnormal point;

(b) if point P_iCharacteristic normal vector N of_i＝(n_ix，n_iy，n_iz) In, n_ixOr n_iyOr n_izIs NaN, then point P_iIs an abnormal point.

Preferably, the sampling module 12 includes a sampling sub-module for removing the abnormal point from the actual workpiece contour point cloud C based on the feature normal vector pair_sAnd workpiece template contour point cloud C_tSampling is carried out, and the method comprises the following steps:

Preferably, the coarse matching module 13 includes:

a maximum workpiece radius obtaining submodule for obtaining a workpiece template C before sampling_tMaximum workpiece radius r_max＝‖P_max-P_minII/2 wherein P_minAnd P_maxFor a workpiece template contour point cloud C_tMinimum point ofP_minAnd a maximum point P_max；

PPF registration submodule for sampling workpiece template contour point cloud C_tsInput point cloud, point cloud C as PPF feature estimator_ssAs target point cloud, 2r_maxAs a distance threshold of the PPF hash map searcher, a PPF registration method is used to obtain a plurality of coarse matching results, namely transformation matrixes; t is_i(i＝0，...，K)；

A rotation-translation submodule for using each transformation matrix T_iPoint-to-point cloud C_tPerforming rotational translation to obtain K transformed point clouds C_ti(i＝0，...，K)；

A template contour point cloud set acquisition submodule for acquiring a point cloud C_tiTurning around an x axis and rotating around a z axis according to conditions; or point cloud C_tiTurning around the y axis and rotating around the z axis according to conditions to respectively obtain a plurality of template contour point cloud sets S_c。

Further, in the template contour point cloud set acquisition submodule, the point cloud C is acquired_tiTurning around the x axis and rotating around the z axis according to conditions to obtain a plurality of template contour point cloud sets S_cThe method comprises the following steps:

z_i＝T_i(0，2，3，1)

s342 the point cloud C_tiOptionally flipped about the x-axis if rotated by the z-axis_iContrary to the vertically downward direction, point cloud C is pointed_tiRotated by 180 ° about the x-axis, which is the transformation matrix T_iX axis of rotation x_iI.e. by

x_i＝T_i(0，0，3，1)

s343 point cloudC_tiRotation around the z-axis conditional: point cloud C_ti(i ═ 0.., K) around the rotational z-axis z_，Rotation n_zEach time of rotation angle is

z_i＝T_i(0，0，1，1)；

s342' Point cloud C_tiOptionally turning around the y-axis if the z-axis is rotated_iContrary to the vertically downward direction, point cloud C is pointed_tRotated by 180 about the x-axis, the y-axis being the transformation matrix T_iX axis y of rotation of_iI.e. by

y_i＝T_i(0，1，3，1)

S343' Point cloud C_tiConditionally rotating about the z-axis, comprising: point cloud C_ti(i ═ 0.., K) around the rotational z-axis z_iRotation n_zEach time of rotation angle is

Preferably, the fine matching module 14 includes:

a matching result obtaining submodule for traversing the set S_cTemplate point cloud C in_jAnd respectively connected with the point cloud C_sPerforming ICP registration, and if the registration is convergent, the matching degree is less than a preset value and the matching proportion is greater than the preset value, determining the point cloud C_jAnd point cloud C_sMatching is successful; obtaining a plurality of matching results which are successfully matched, including a transformation matrix T_jMatching degree and matching proportion; and if the matching result of the matching is not successful, the 3D template matching is considered to be failed.

A matching result sorting module for sorting the matching results according to the matching proportion and outputting the transformation matrix of the optimal matching result as T_b。

EXAMPLE III

Fig. 8 is a schematic structural diagram of a computer device according to an embodiment of the present invention, such as a smart phone, a tablet computer, a notebook computer, a desktop computer, a rack-mounted server, a blade server, a tower server, or a rack-mounted server (including an independent server or a server cluster formed by multiple servers) that can execute programs. The computer device 20 of the present embodiment includes at least, but is not limited to: a memory 21, a processor 22, which may be communicatively coupled to each other via a system bus, as shown in FIG. 8. It is noted that fig. 8 only shows a computer device 20 with components 21-22, but it is to be understood that not all shown components are required to be implemented, and that more or fewer components may be implemented instead.

In this embodiment, the memory 21 (i.e., the readable storage medium) includes a Flash memory, a hard disk, a multimedia Card, a Card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a Read Only Memory (ROM), an Electrically Erasable Programmable Read Only Memory (EEPROM), and a Programmable Read Only Memory (PROM), and the memory 21 may also be an external storage device of the computer device 20, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash Card (Flash Card), and the like provided on the computer device 20. Of course, the memory 21 may also include both internal and external storage devices of the computer device 20. In this embodiment, the memory 21 is generally used for storing an operating system and various types of application software installed in the computer device 20, such as program codes of the template matching apparatus in the method embodiment. Further, the memory 21 may also be used to temporarily store various types of data that have been output or are to be output.

Processor 22 may be a Central Processing Unit (CPU), controller, microcontroller, microprocessor, or other data Processing chip in some embodiments. The processor 22 is typically used to control the overall operation of the computer device 20. In this embodiment, the processor 22 is configured to execute the program code stored in the memory 21 or process data, for example, execute the template matching apparatus 11, so as to implement the template matching method in the method embodiment.

Example four

The present application also provides a computer-readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, a server, an App application mall, etc., on which a computer program is stored, which when executed by a processor implements corresponding functions. The computer-readable storage medium of the present embodiment is used for storing program codes of a template matching apparatus, and when executed by a processor, implements the template matching method in the method embodiment.

It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflict.

The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.

Claims

1. A method for matching a flat workpiece template is characterized by comprising the following steps:

S3 aligning the point cloud C_tsAnd the point cloud C_ssCarrying out coarse matching positioning to obtain a plurality of coarse matching transformation matrixes T_i(i ═ 0.. times, K), and based on each of the transformation matrices T_iFor the point cloud C_tPerforming rotational translation to obtain K point clouds C after transformation_ti(i ═ 0.., K); k point clouds C_tiOptionally rotating around x-axis and z-axis, or adding K point clouds C_tiTurning around the y axis and rotating around the z axis according to conditions to obtain a plurality of template contour point clouds S_c；

2. The method for matching a flat workpiece template according to claim 1, wherein in the step S2, the feature normal vector acquisition comprises the following steps: obtaining the geometric center point P of the point cloud C_centexGo through each point P of the point cloud C_iFor each point P_iIts characteristic normal vector

3. The method of claim 1, wherein in step S2, the point cloud C is obtained_tAnd point cloud C_sPoint P in_iA point which is considered abnormal when any one of the following conditions is satisfied;

4. The method of claim 1, wherein in step S2, the point cloud C of the contour of the actual workpiece is obtained after removing the abnormal points based on the pairs of feature normal vectors_sAnd workpiece template contour point cloud C_tThe sampling comprises the following steps:

For addition to C_oPoint V in_iSynchronously performing step (c) comprising: at a predetermined distance d₀Search point within range V_iNeighbor point of (V)_i1，V_i2，...V_imM denotes a point V_iAnd judging V_iIs a neighbor point V_ijWhether a preset constraint condition is met or not, wherein the preset condition is as follows: neighbor point V_ijUntreated and V_iAnd V_ijAre not more than preSetting the angle threshold theta₀；

If V_iIts neighbor point V_iiAfter processing, skipping point V_iiJudgment of V_iWhether other neighbor points of (1) satisfy the constraint condition;

5. The method for matching a flat workpiece template as claimed in claim 1, wherein in step S3, the point cloud C is registered by using a PPF registration method_tsAnd point cloud C_ssPerforming coarse matching positioning, including:

s31 acquiring workpiece template C before sampling_tMaximum workpiece radius r_max＝||P_max-P_minI/2, wherein P_minAnd P_maxFor a workpiece template contour point cloud C_tMinimum point P of_minAnd a maximum point P_max；

S34 point cloud C_tiConditional wrap transform matrix T_iX-axis flipping and z-axis rotation; or point cloud C_tiOptionally rotating around the y-axis and the z-axisObtaining a plurality of template contour point cloud sets S_c。

6. A method as claimed in claim 1 or 5, characterized in that the point cloud C is used as a template for matching planar workpieces_tiTurning around the x axis and rotating around the z axis according to conditions to obtain a plurality of template contour point cloud sets S_cThe method comprises the following steps:

z_i＝T_i(0，2，3，1)

x_i＝T_i(0，0，3，1)

s343 point cloud C_tiRotation around the z-axis conditional: point cloud C_ti(i ═ 0.., K) around the rotational z-axis z_iRotation n_zEach time of rotation angle is

or, the point cloud C_tiConditional wrap transform matrix T_iThe method comprises the following steps of turning over the y axis and rotating the z axis to obtain a plurality of template contour point clouds:

s341' from the point cloud C_ti(i ═ 0.... K.) corresponding transformation matrix T_iCalculate the pair thereofZ axis of rotation z_i：

z_i＝T_i(0，0，1，1)；

s342' Point cloud C_tiTurning around the y axis according to conditions: if the z-axis zi is rotated in a direction opposite to the vertical downward direction, the point cloud C is pointed_tiRotated by 180 about the x-axis, the y-axis being the transformation matrix T_iX axis y of rotation of_iI.e. by

y_i＝T_i(0，1，3，1)

Wherein, T_i(0, 0, 3, 1) represents a number from T_iThe number of the 0 th row and the 1 st column is 3, so that 1 column is formed;

7. The method of claim 1, wherein in step S4, ICP is used to set S_cTemplate point cloud and workpiece contour point cloud C in_sPerforming fine matching positioning, including:

s41 traversing set S_cTemplate point cloud C in_jAnd respectively connected with the point cloud C_sPerforming ICP registration, and if the registration is convergent, the matching degree is less than a preset value and the matching proportion is greater than the preset value, determining the point cloud C_jAnd point cloud C_sMatching is successful; obtaining a plurality of matching results which are successfully matched, including a transformation matrix T_jMatching degree and matching proportion; if the matching result of the matching is not successful, the 3D template matching is considered to be failed;

s42, according to the matching proportion, the matching results in the step S41 are sorted, and the transformation matrix of the output optimal matching result is recorded as T_b。

8. The method as claimed in claim 1, wherein in step S1, the actual workpiece contour point cloud C is obtained_sThe method comprises the following steps:

obtaining actual workpiece point cloud C₁Based on a preset rule, the actual workpiece point cloud C is processed₁Conversion into a tetragonal grid SM₁And acquire the SM₁Normal vector of each vertex in the vector;

segmenting the SM based on the normal vector and a preset constraint condition₁Removing the SM based on the segmentation result₁Generating a four-corner grid SM after removing the side point cloud₂；

Searching for and acquiring the SM₂All the edge points of (a), said edge points constituting a point cloud C₃The actual contour point cloud of the workpiece to be extracted is obtained;

and/or the workpiece template contour point cloud C_tThe acquisition comprises the following steps:

analyzing a 2D drawing of a workpiece template, and converting a straight line entity, an arc entity and a circle entity of the 2D drawing into a line data structure, an arc data structure and a circle data structure according to preset rules respectively to obtain a plurality of template reference lines S_i(i＝1，2，..，N_s)，N_sRepresenting the total number of the reference lines;

analyzing all the line data structure, the arc data structure and the circle data structure according to a given resolution l₀Respectively dispersing to generate corresponding reference line point clouds;

and merging all the reference line point clouds to obtain the workpiece template point cloud.

9. A planar workpiece template matching apparatus, comprising:

a point cloud obtaining module for obtaining the actual workpiece contour point cloud C_sHe Gong workerPiece template contour point cloud C_t；

A fine matching module for matching a plurality of the template contour point clouds S_cAnd the actual workpiece contour point cloud C_sCarrying out fine matching positioning, and outputting a transformation matrix of an optimal matching result to be recorded as T_b。

10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method of any of claims 1 to 8 are implemented by the processor when executing the computer program.

11. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 8.