Point cloud segmentation method and device for saddle-shaped weld groove of thick plate and electronic equipment

Abstract

The invention belongs to the technical field of weld joint recognition, and particularly discloses a point cloud segmentation method and device for a saddle-shaped weld joint groove of a thick plate and electronic equipment. The method comprises the following steps: carrying out dimension reduction processing on obtained three-dimensional point cloud data of a saddle-shaped weld groove to obtain target plane point cloud data; performing position numbering on each point in the target plane point cloud data, and determining a plurality of segmentation points from the numbered target plane point cloud data so as to determine a plurality of segmentation segments; iteratively optimizing the selection position of each segmentation point by taking minimization of the target function as a target until each optimal segmentation point is determined; the objective function is determined based on the endpoint fitting deviation and the slope consistency deviation of each segmentation section; and segmenting the three-dimensional point cloud data according to each optimal segmentation point. According to the invention, the point cloud segmentation precision of the saddle-shaped weld groove can be greatly improved, and the point cloud data of different parts of the saddle-shaped weld groove can be accurately extracted.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of weld seam recognition, and more particularly, to a thick plate saddle-shaped weld seam groove point cloud segmentation method and device and electronic equipment. BACKGROUND

[0002] At present, the weld seam groove of laser vision recognition is mainly straight weld seams of single-layer single-pass, single-layer multi-pass, fillet welds, and the like. Such weld seam grooves present a v-shaped or I-shaped structure with left-right symmetry. In the prior art, by setting the number of point clouds in each part, the segmentation of the point cloud data of the above-mentioned regular weld seam groove can be effectively realized by using a simple linear fitting method to identify weld defects.

[0003] However, in actual industrial production, there are often multi-layer multi-pass variable cross-section thick plate saddle-shaped weld seam grooves that are spatially intersected. Such weld seam grooves present a spatially varying curve, have a complex geometric shape and spatial diversity, the opening direction of the groove is variable, and the cross-section presents an irregular contour. Under the combined influence of these factors, the segmentation accuracy is low when the above-mentioned existing segmentation algorithm is used to segment the three-dimensional point cloud data of the saddle-shaped weld seam groove, and effective segmentation is difficult to achieve.

[0004] Therefore, how to effectively realize the point cloud segmentation of the thick plate saddle-shaped weld seam groove has become a technical problem to be solved in the industry. SUMMARY

[0005] In view of the defects of the prior art, the purpose of the present application is to effectively realize the point cloud segmentation of the thick plate saddle-shaped weld seam groove, and to solve the problem of low segmentation accuracy and difficulty in effective segmentation when the existing segmentation algorithm is used to segment the three-dimensional point cloud data of the saddle-shaped weld seam groove.

[0006] To achieve the above-mentioned purpose, in a first aspect, the present application provides a thick plate saddle-shaped weld seam groove point cloud segmentation method, comprising: performing dimension reduction processing on the acquired three-dimensional point cloud data of the saddle-shaped weld seam groove to obtain target planar point cloud data; numbering each point in the target planar point cloud data, and determining a plurality of segmentation points from the numbered target planar point cloud data to determine a plurality of segmentation sections; iteratively optimizing the selection position of each segmentation point to minimize the target function until the optimal segmentation points are determined; the target function is determined based on the endpoint fitting deviation and the slope consistency deviation of each segmentation section; segmenting the three-dimensional point cloud data according to each optimal segmentation point.

[0007] Optionally, the iterative optimization of the selection position of each segmentation point to minimize the target function until the optimal segmentation points are determined comprises: Step S101, under the preset position constraint condition, the selected position of each of the segmentation points in the current iteration process is adjusted according to a preset displacement strategy, and the function value of the objective function corresponding to the selected position of each of the segmentation points after adjustment is calculated; Step S102, the optimal displacement amount corresponding to the minimum objective function value in each of the function values is determined; Step S103, according to the selected position of each of the segmentation points in the current iteration process and the optimal displacement amount, each of the segmentation points in the next iteration process is determined. Steps S101 to S103 are repeated until the function value of the objective function meets the preset convergence condition, and each optimal segmentation point is determined.

[0008] Optionally, the preset displacement strategy is determined based on equal division of a step function in the current iteration process; the step function is determined based on the total number of point cloud data of the target plane and a decay factor in the current iteration process, and the decay factor is determined according to the minimum objective function value in the last iteration process and the initial function value of the objective function.

[0009] Optionally, before the method of minimizing the objective function to iteratively optimize the selected position of each of the segmentation points until each optimal segmentation point is determined, the method further comprises: Based on the coordinate information of each of the segmentation points and the coordinate information of each point in each of the segmentation segments, the endpoint fitting deviation and the slope consistency deviation of each of the segmentation segments are determined. Based on the endpoint fitting deviation and the slope consistency deviation of each of the segmentation segments, and a preset segment length penalty term parameter, the objective function is determined.

[0010] Optionally, the determination step of the endpoint fitting deviation of each of the segmentation segments specifically comprises: Based on the coordinate information of each of the segmentation points, a linear fitting function of each of the segmentation segments is determined. The amplitude and the number of points in each of the segmentation segments are determined. Based on the coordinate information of each point in each of the segmentation segments, the linear fitting function, the amplitude, and the number of points in each of the segmentation segments, the endpoint fitting deviation of each of the segmentation segments is determined.

[0011] Optionally, the determination step of the slope consistency deviation of each of the segmentation segments specifically comprises: Based on the coordinate information of each of the segmentation points, the endpoint fitting slope of each of the segmentation segments is determined. Based on the coordinate information of each point in each of the segmentation segments, the least square slope of each of the segmentation segments is determined. A slope consistency deviation of each of the segmentation segments is determined based on a slope fitted at an end point of each of the segmentation segments and a least square slope.

[0012] Optionally, the acquired three-dimensional point cloud data of the saddle-shaped weld groove is subjected to dimension reduction processing to obtain target plane point cloud data, including: The three-dimensional point cloud data is subjected to point cloud coordinate decentralization processing to obtain decentralized three-dimensional point cloud data. The target rotation matrix is determined according to the decentralized three-dimensional point cloud data. The three-dimensional point cloud data is projected onto a target plane according to the target rotation matrix to obtain the target plane point cloud data.

[0013] In a second aspect, the present application provides a thick plate saddle-shaped weld groove point cloud segmentation device, including: A dimension reduction module is configured to subject the acquired three-dimensional point cloud data of the saddle-shaped weld groove to dimension reduction processing to obtain target plane point cloud data. A first processing module is configured to number positions of each point in the target plane point cloud data and determine a plurality of segmentation points from the numbered target plane point cloud data to determine a plurality of segmentation segments. A second processing module is configured to iteratively optimize a selected position of each of the segmentation points until optimal segmentation points are determined, with a minimized target function as a target; the target function is determined based on a fitting deviation of an end point of each of the segmentation segments and a slope consistency deviation. A segmentation module is configured to segment the three-dimensional point cloud data according to the optimal segmentation points.

[0014] In a third aspect, the present application provides an electronic device, including at least one memory configured to store a program, and at least one processor configured to execute the program stored in the memory, when the program stored in the memory is executed, the processor is configured to execute the method described in the first aspect or any possible implementation manner of the first aspect.

[0015] In a fourth aspect, the present application provides a computer readable storage medium, and the computer readable storage medium stores a computer program, when the computer program is run on a processor, the processor is caused to execute the method described in the first aspect or any possible implementation manner of the first aspect.

[0016] In a fifth aspect, the present application provides a computer program product, when the computer program product is run on a processor, the processor is caused to execute the method described in the first aspect or any possible implementation manner of the first aspect.

[0017] It can be understood that the beneficial effects of the above-mentioned second aspect to the fifth aspect can be referred to the related description in the above-mentioned first aspect, which will not be repeated here.

[0018] Overall, compared with the prior art, the above technical solutions conceived by the present application have the following beneficial effects: The thick plate saddle-shaped weld groove point cloud segmentation method, device and electronic equipment provided by the present application can obtain spatially intersecting multi-layer multi-pass variable cross-section saddle-shaped weld groove three-dimensional point cloud data through online laser vision frame-by-frame scanning, convert the point cloud data to a target plane for data dimensionality reduction analysis, divide the point cloud data into multiple segmentation sections according to the spatial variation curve characteristics contained in the saddle-shaped weld groove, and perform linear fitting on each segmentation section. Meanwhile, the deviation and slope consistency measurement are estimated by combining the fitting straight line of the end points of the segmentation sections, a composite objective function is constructed for multi-objective optimization solving, the identification problem of the irregular contour of the cross section of the saddle-shaped weld groove is overcome, the point cloud segmentation accuracy for the saddle-shaped weld groove can be greatly improved, and the point cloud data of different parts of the saddle-shaped weld groove can be accurately extracted. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a flowchart of the thick plate saddle-shaped weld groove point cloud segmentation method provided by the embodiments of the present application; Figure 2 is a result diagram of the dimensionality reduction of the saddle-shaped weld groove point cloud data provided by the embodiments of the present application; Figure 3 (a) in is a real scene diagram of laser scanning point cloud data acquisition of the thick plate saddle-shaped weld groove provided by the embodiments of the present application, and (b) is a diagram of feature extraction of the thick plate saddle-shaped weld groove point cloud data provided by the embodiments of the present application; Figure 4 (a) in is a diagram of the point cloud data of a variable cross-section saddle-shaped weld groove on the plane provided by the embodiments of the present application, and (b) is a diagram of the point cloud data of another variable cross-section saddle-shaped weld groove on the plane provided by the embodiments of the present application; x - z (a) in is a diagram of the point cloud data of a variable cross-section saddle-shaped weld groove on the plane provided by the embodiments of the present application, and (b) is a diagram of the point cloud data of another variable cross-section saddle-shaped weld groove on the plane provided by the embodiments of the present application; x - z (a) in is a diagram of the point cloud data of a variable cross-section saddle-shaped weld groove on the plane provided by the embodiments of the present application, and (b) is a diagram of the point cloud data of another variable cross-section saddle-shaped weld groove on the plane provided by the embodiments of the present application; Figure 5 is a structural diagram of the thick plate saddle-shaped weld groove point cloud segmentation device provided by the embodiments of the present application; Figure 6 is a structural diagram of the electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0020] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application and not to limit the present application.

[0021] The terms "first" and "second" and the like in the specification and claims of the present application are used to distinguish different objects and are not used to describe a specific order of the objects. For example, the first processing module and the second processing module are used to distinguish different processing modules and are not used to describe a specific order of the processing modules.

[0022] In the embodiments of the present application, the words such as "exemplary" or "for example" are used to mean serving as an example, illustration, or description. Any embodiment or design solution described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or having more advantages than other embodiments or design solutions. Rather, the words such as "exemplary" or "for example" are intended to present the relevant concept in a specific manner.

[0023] The embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.

[0024] Figure 1 is a flowchart of a thick plate saddle-shaped weld groove point cloud segmentation method provided by the embodiments of the present application, as shown in Figure 1 , comprising: Step S1, performing dimension reduction processing on the acquired three-dimensional point cloud data of the saddle-shaped weld groove to obtain target plane point cloud data; Step S2, performing position numbering on each point in the target plane point cloud data, and determining a plurality of segmentation points from the numbered target plane point cloud data to determine a plurality of segmentation sections; Step S3, iteratively optimizing the selection position of each segmentation point to determine each optimal segmentation point, with the goal of minimizing the target function; the target function is determined based on the endpoint fitting deviation and the slope consistency deviation of each segmentation section; Step S4, segmenting the three-dimensional point cloud data according to each optimal segmentation point.

[0025] Specifically, the target plane point cloud data described in the embodiments of the present application refers to the point cloud data obtained by projecting the three-dimensional point cloud data of the saddle-shaped weld groove onto a target plane. The target plane can be set according to the direction of laser scanning, and it is only necessary to ensure that the target plane point cloud data contains the height information and the overall shape of the saddle-shaped weld groove, so that each segmentation is targeted at the point cloud data of a single laser scan.

[0026] The segmentation point described in the embodiments of the present application refers to a point in the numbered target plane point cloud data for segmenting the entire target plane point cloud data.

[0027] In other words, after determining each segmentation point, the entire target planar point cloud data can be segmented according to the position number of each segmentation point. Thus, the entire point cloud covered by the first point to the first segmentation point (inclusive) in the numbered target planar point cloud data can be considered the first segment; the entire point cloud covered by the first segmentation point to the second segmentation point (inclusive) can be considered the second segment; the entire point cloud covered by the second segmentation point to the third segmentation point (inclusive) can be considered the third segment, and so on, dividing the entire target planar point cloud data into multiple segments.

[0028] In embodiments of this application, each point in the point cloud data of the target plane can be numbered according to the saddle-shaped weld bevel along the weld extension direction on the target plane. For example... Figure 2 As shown, through spatial dimensionality reduction processing, the target plane point cloud data is projected into a 3D point cloud onto... x - z Point cloud data on a plane, point cloud in the plane y The values ​​are the same. All points in the target planar point cloud data can be classified according to... x Numbering is done along the axis direction, where for the same axis... x The value point, inside can be arranged along z By numbering along the axis, the location of each point in the target planar point cloud data can be assigned.

[0029] The objective function described in this application embodiment can be determined based on the endpoint fitting deviation and slope consistency deviation of each segment. The endpoint fitting deviation characterizes the average distance deviation from the fitted curve of the line connecting each point within each segment to the endpoint; the slope consistency deviation characterizes the slope deviation between the endpoint fitting slope of each segment and the least squares curve. Specifically, the endpoint fitting deviation and slope consistency deviation can be calculated based on the coordinate information of each segmentation point and the coordinate information of each point within each segment.

[0030] The optimal segmentation point described in the embodiments of this application refers to the segmentation point determined by minimizing the objective function.

[0031] In the embodiments of this application, in step S1, three-dimensional point cloud data of the saddle-shaped weld groove is acquired by three-dimensional laser scanning. Assume the three-dimensional point cloud dataset of the saddle-shaped weld groove is... It can be represented as: (1) In the formula, represents the saddle-shaped weld bevel groove profile point cloud data obtained by the jth frame of laser vision scanning. Wherein, , N represents the total number of points when scanning a frame of profile, and M represents the total number of frames of three-dimensional point cloud of the saddle-shaped weld bevel groove obtained by laser vision scanning.

[0032] By spatial dimension reduction processing, the above three-dimensional point cloud can be projected onto x - z a plane to obtain target plane point cloud data, that is, the saddle-shaped weld bevel groove point cloud data set .

[0033] Further, in step S2, each point in the target plane point cloud data is numbered by position, which can be numbered from 1 to x according to the aforementioned N axis direction numbering method, and then a plurality of initialized segmentation points can be determined from the numbered target plane point cloud data, so as to determine a plurality of segmentation sections. The next goal is to seek an optimal segmentation point sequence to divide the data into K = continuous sub-sections, that is, four segmentation sections can be obtained. This process can be represented as: ;(2) ;(3) Here, represents the first point in the target plane point cloud data.

[0034] Wherein, each segmentation section covers point cloud data satisfying the linear assumption: ;(4) It should be noted that the saddle-shaped weld bevel groove is left-right asymmetric, and its groove profile can be composed of four parts of branch pipe outer wall, groove side wall, weld backing surface and main pipe outer wall. The surfaces of the branch pipe and the main pipe are the surfaces of circular pipes with different ellipticities. After artificial backing and polishing, the backing surface is uneven and does not have the symmetry characteristic. Therefore, the number of segmentation points k=1, 2, 3 can be set; the number of segmentation sections K =1, 2, 3, 4.

[0035] Further, in the embodiments of the present application, in step S3, the endpoint fitting deviation and the slope consistency deviation of each segmentation segment are combined to construct a target function, and then the selection position of each segmentation point is iteratively optimized until the optimal segmentation points are determined. Specifically, based on the statistical characteristics of point cloud differential geometry, the point cloud data is segmented into multiple segments by a recursive method, and the point cloud data in each segment has similar differential geometric statistical characteristics. The data is segmented and it is assumed that each segment of data has a linear relationship. The slope and fitting error of each segmentation segment are calculated by constructing a target function. The selection position of each segmentation point is iteratively optimized. The linear fitting parameters of adjacent segmentation segments are compared to identify the change point. When the deviation exceeds the set threshold, it is determined that there is a change point. Finally, the position of the change point is returned to determine the optimal segmentation point.

[0036] Further, in the embodiments of the present application, in step S4, the three-dimensional space mapping relationship of each frame of laser scanning point cloud data is utilized, and according to the optimal segmentation points, the corresponding optimal segmentation points in each frame of laser scanning point cloud data can be correspondingly determined. Then, by inter-frame combination, the optimal segmentation points in each frame of point cloud data are correspondingly connected, so that the accurate segmentation of the saddle-shaped weld groove three-dimensional point cloud data can be realized. This method can more stably and accurately process different saddle-shaped weld groove three-dimensional point cloud data under different scanning postures and different underlying polishing conditions, and realize effective point cloud segmentation and point cloud data extraction of the branch pipe outer wall part, the groove side wall part, the weld underlying surface and the main pipe outer wall.

[0037] It should be noted that the form of each segment of the weld groove after three-dimensional point cloud segmentation and the number of point clouds contained are related to the geometric profile characteristics, spatial position and laser vision scanning angle of the saddle-shaped weld groove.

[0038] In actual production, the TKY type pipe joint intersection weld is significantly different from the traditional weld in geometric appearance. The groove cross-sectional profile is not a regular straight line or circular curve, but a combination of straight lines, elliptical curves and irregular curves. Figure 3 As shown in (a) of FIG. 1, Figure 3 (a) is a real scene schematic diagram of laser scanning point cloud data acquisition of a thick plate saddle-shaped weld groove provided by the embodiments of the present application, and (b) is a schematic diagram of feature extraction of thick plate saddle-shaped weld groove point cloud data provided by the embodiments of the present application. The groove cross-sectional profile is composed of laser projections of the branch pipe outer wall, the groove side wall, the main pipe outer wall and the weld underlying surface. When the laser irradiates the groove side wall plane, the projection is a straight line profile. The branch pipe outer wall and the inner wall profile are elliptical profiles due to the laser irradiation on the circular pipe at a certain angle. The underlying weld surface profile presents an irregular profile due to the irregularity of the underlying surface.

[0039] The thick plate saddle-shaped weld groove point cloud segmentation method of the embodiment of the application obtains the spatial intersecting multi-layer multi-pass variable cross-section saddle-shaped weld groove three-dimensional point cloud data through online laser vision frame-by-frame scanning, converts the point cloud data to a target plane for data dimension reduction analysis, divides the point cloud data into multiple segmentation sections according to the spatial variation curve characteristics contained in the saddle-shaped weld groove, and performs linear fitting on each segmentation section, while using the fitting straight line of the end points of the segmentation sections to estimate the consistency of the deviation and slope, combining the two to construct a composite objective function for multi-objective optimization solving, thereby overcoming the identification problem of the irregular contour of the cross section of the saddle-shaped weld groove, greatly improving the point cloud segmentation accuracy for the saddle-shaped weld groove, and accurately extracting the point cloud data of different parts of the saddle-shaped weld groove.

[0040] Based on the content of the above embodiment, as an optional embodiment, step S1, the three-dimensional point cloud data of the acquired saddle-shaped weld groove is subjected to dimension reduction processing to obtain target plane point cloud data, including: The three-dimensional point cloud data is subjected to point cloud coordinate decentralization processing to obtain decentralized three-dimensional point cloud data; The decentralized three-dimensional point cloud data is subjected to singular value decomposition to determine a target rotation matrix; The three-dimensional point cloud data is projected onto the target plane according to the target rotation matrix to obtain the target plane point cloud data.

[0041] Specifically, in the embodiment of the application, the acquired saddle-shaped weld groove three-dimensional point cloud data is clean data after denoising and filtering processing of the original point cloud data. When the three-dimensional point cloud data is subjected to dimension reduction processing, the singular value decomposition method can be used to project the three-dimensional point cloud data onto the target plane.

[0042] More specifically, the saddle-shaped weld groove contour point cloud data obtained by the jth frame of laser vision scanning is assumed to be The corresponding centralized point cloud coordinates are Then, (5) The three-dimensional point cloud coordinates are subjected to decentralization processing, and then (6) The decentralized three-dimensional point cloud data .

[0043] Wherein .

[0044] Because the singular value decomposition (SVD) method has high stability and strong noise resistance in processing point cloud data distribution characteristics, this method can be used to extract the three-dimensional point cloud plane normal vectors of the saddle-shaped weld groove, providing basic data for subsequent point cloud coordinate transformation.

[0045] Further, the decentralized three-dimensional point cloud SVD decomposition is performed, as shown in equation (7), and the left singular vector corresponding to the smallest singular value is the normal vector of the plane, that is: ; (7) wherein, is an orthogonal matrix of , is a diagonal matrix of , is an orthogonal matrix of ; the plane normal vector is determined according to the matrix , that is, . .

[0046] Further, the rotation axis and the rotation angle are calculated, as shown in equations (8) and (9).

[0047] ; (8) ; (9) wherein, .

[0048] The expression of the rotated three-dimensional point cloud is shown in equations (10)-(12), and the target rotation matrix R is determined, and the target plane point cloud data is determined according to the target rotation matrix R, that is: ; (10) ; (11) ; (12) The method of the embodiment can project the three-dimensional point cloud to the target plane by using the rotation matrix after the original point cloud data is filtered and denoised, realize dimension reduction of the three-dimensional point cloud to the two-dimensional space, reduce the feature analysis dimension under the premise of ensuring the retention of the weld contour geometric information, and significantly reduce the processing difficulty and the calculation amount.

[0049] Based on the above-mentioned embodiment, as an optional embodiment, in step S3, the selected position of each segmentation point is iteratively optimized to minimize the target function, and before the optimal segmentation point is determined, the method further includes: Based on the coordinate information of each segmentation point and the coordinate information of each point in each segmentation section, the endpoint fitting deviation and the slope consistency deviation of each segmentation section are determined. ​The target function is determined based on the endpoint fitting deviation and the slope consistency deviation of each segmentation segment, and a preset segmentation length penalty term parameter.

[0050] Specifically, in the embodiments of the present application, in step S3, the selected positions of each segmentation point are iteratively optimized with the objective of minimizing the target function, and before determining the respective optimal segmentation points, the target function needs to be constructed.

[0051] In the embodiments of the present application, for a complex welding groove with a spatial variable cross-section, a fitting method combining the endpoint fitting deviation and the slope consistency measure is introduced, after the initial determination of each segmentation point from the numbered target plane point cloud data, the coordinate information of each segmentation point and the coordinate information of each point in each segmentation segment are used to determine the endpoint fitting deviation and the slope consistency deviation of each segmentation segment.

[0052] Based on the content of the above embodiments, as an optional embodiment, the determination step of the endpoint fitting deviation of each segmentation segment specifically includes: determining a linear fitting function of each segmentation segment based on the coordinate information of each segmentation point; determining the amplitude and the number of points in each segmentation segment; determining the endpoint fitting deviation of each segmentation segment according to the coordinate information of each point in each segmentation segment, the linear fitting function, the amplitude and the number of points in each segmentation segment.

[0053] Specifically, in the embodiments of the present application, the specific calculation implementation of the endpoint fitting deviation of each segmentation segment can be obtained by formula (13): For the candidate segmentation point Define the endpoint fitting deviation of the Kth segment: ; (13) In the formula, represents the number of points in the segmentation segment; represents the amplitude of the segmentation segment; represents the linear fitting function of the segmentation segment.

[0054] wherein, ; ; ; Therefore, the coordinate information of each segmentation point and the coordinate information of each point in each segmentation segment can be substituted into the above formulas to determine the function value, amplitude and number of points in each segmentation segment of the linear fitting function, and then the endpoint fitting deviation of each segmentation segment can be calculated by formula (13) .

[0055] The method of the embodiment of the present application divides the target planar point cloud data into multiple segments according to the geometric characteristics of the saddle-shaped weld groove, determines the endpoint fitting deviation of each segment by using the coordinate information of each point in each segment, the linear fitting function of each segment, the amplitude, and the number of points in the segment, and ensures the accuracy of the endpoint fitting deviation calculation.

[0056] Based on the content of the above embodiment, as an optional embodiment, the determination of the slope consistency deviation of each segment specifically includes: determining the endpoint fitting slope of each segment based on the coordinate information of each segment point; determining the least square slope of each segment based on the coordinate information of each point in each segment; determining the slope consistency deviation of each segment based on the endpoint fitting slope and the least square slope of each segment.

[0057] Specifically, in the embodiment of the present application, the specific calculation implementation of the slope consistency deviation of each segment can be obtained by formula (14): (14) In the formula: the endpoint fitting slope of the segment, the least square slope of the segment.

[0058] wherein, ; ; It can be understood that the endpoint fitting slope refers to the slope calculated by linear fitting using the two end points of the segment; and the least square slope refers to the slope calculated by least square fitting using all the point clouds in the segment.

[0059] Therefore, the coordinate information of each segment point and the coordinate information of each point in each segment can be substituted into the above formulas to determine the endpoint fitting slope and the least square slope of each segment, and then the slope consistency deviation of each segment can be calculated by formula (14) .

[0060] The method of the embodiment of the present application divides the target planar point cloud data into multiple segments according to the geometric characteristics of the saddle-shaped weld groove, determines the endpoint fitting slope and the least square slope of each segment by using the coordinate information of each segment point and the coordinate information of each point in each segment, determines the slope consistency deviation of each segment by combining the endpoint fitting slope and the least square slope, and ensures the accuracy of the slope consistency deviation calculation.

[0061] Further, in the implementation of the present application, a segment length penalty term can be introduced, and based on the endpoint fitting deviation and the slope consistency deviation of each segmentation segment, as well as a preset segment length penalty term parameter, a target function is constructed.

[0062] Specifically, the target function after combining the multi-target and introducing the segment length penalty term can be expressed as: (15) wherein, denotes a segmentation point vector; denotes a weight factor; denotes a length penalty coefficient; denotes a decay parameter Here, in the embodiments of the present application, the single segment target quasi-convexity of the target function is also proved, as described below.

[0063] Fixing and , the function is quasi-convex on the interval .

[0064] The proof process is as follows: Let , then: ; ; When approaches the true segmentation point, the derivative changes sign and is zero at the true value.

[0065] The method of the embodiments of the present application combines the endpoint fitting deviation and the slope consistency measurement, designs a composite target function with length penalty, and designs an efficient optimization algorithm using the quasi-convexity of the target function, so as to facilitate the rapid positioning of the turning region of the saddle-shaped weld bevel point cloud and the point cloud segmentation.

[0066] Based on the content of the above embodiments, as an optional embodiment, step S3 aims to minimize the target function, and iteratively optimizes the selected position of each segmentation point until the optimal segmentation points are determined, including: Step S101, under the preset position constraint condition, the selected position of each segmentation point in the current iteration process is adjusted according to a preset displacement strategy, and the function value of the target function corresponding to the selected position of each segmentation point after adjustment is calculated; Step S102, determining the optimal displacement amount corresponding to the minimum target function value in each function value; Step S103, determining each segmentation point in the next iteration process according to the selected position of each segmentation point and the optimal displacement amount; Steps S101 to S103 are repeated until the function value of the target function meets a preset convergence condition, and each optimal segmentation point is determined.

[0067] Specifically, in the embodiments of the present application, each segmentation point is adaptively optimized. Specifically, for each segmentation point, several candidate positions (including the current point) are taken near the current point, the target function value is calculated, and then the position that minimizes the target function is selected as the new position of the segmentation point. Repeat this process until the target function converges.

[0068] More specifically, let denote the segmentation point vector of the i-th iteration, n be the target function defined on the segmentation point space, denote a 1*3 positive integer vector. The algorithm solves the following optimization problem by coordinate rotation method: ; (16) In the embodiments of the present application, the position of the segmentation point is first initialized. Specifically, the initial segmentation point can be uniformly distributed, that is, ; Based on the determination of the initial segmentation point, step S101 is performed, that is, under the preset position constraint condition, the selected position of each segmentation point in the current iteration process is adjusted according to the preset displacement strategy, and the function value of the target function corresponding to the selected position of each segmentation point after adjustment is calculated.

[0069] Based on the content of the above embodiments, as an optional embodiment, the preset displacement strategy is determined based on equal division of the step function in the current iteration process; the step function is determined based on the total number of point cloud data of the target plane and the decay factor in the current iteration process, and the decay factor is determined according to the minimum target function value in the last iteration process and the initial function value of the target function.

[0070] Specifically, in the embodiments of the present application, an adaptive step strategy is adopted, and the adaptive step function is defined as: ; (17) wherein, denotes the decay factor, ; the adaptive step parameter , N denotes the total number of point cloud data of the target plane.

[0071] Here, denotes the initial function value of the target function, that is, . ​

[0072] Further, the preset displacement strategy is determined based on equal division of the step function in the current iteration process. Specifically, the preset displacement strategy is determined as follows: (18) wherein, is a displacement amount, representing an adjustment amplitude of the current segmentation point position . is a candidate displacement amount set.

[0073] In the embodiments of the present application, the update strategy of coordinate rotation can be used. For each segmentation point t, the segmentation point position is adjusted by a preset displacement amount. Through the discrete displacement trial, the optimal segmentation point position is found under the premise of guaranteeing the topological constraint.

[0074] The method of the embodiments of the present application avoids the step sensitivity problem of the continuous optimization method (such as gradient descent) by using the above-mentioned displacement strategy design, while maintaining a relatively low computational complexity, improving the optimization efficiency and accuracy of the algorithm, and realizing the adaptive point cloud segmentation of the saddle-shaped weld groove.

[0075] Further, for each segmentation point, the preset position constraint condition is determined as follows: (19) wherein, is a minimum segment length constraint, .

[0076] Here, it should be noted that is a trial step, which is used to perform local search in the vicinity of the current segmentation point to traverse each displacement amount in the candidate displacement set . The candidate position after displacement for each is . Through the above-mentioned preset position constraint condition , it can be ensured that the selected position of the segmentation point after displacement satisfies the maximum change point number constraint, i.e. is not less than the position of the previous segmentation point plus the minimum segment length, and is not greater than the position of the next segmentation point minus the minimum segment length.

[0077] By evaluating each , the selected position of each current segmentation point is adjusted, and the function value of the objective function corresponding to the selected position of each segmentation point after adjustment is calculated .

[0078] Further, in step S102, the optimal displacement amount corresponding to the minimum objective function value in each function value can be determined according to formula (20): (20) in, , which represents the state of fixing other dividing points.

[0079] For example, the n Starting from the next iteration, optimize hour, That is, fixed , ,Sure Furthermore, optimization At that time, it needs to be fixed. , , determine Furthermore, optimize At that time, it needs to be fixed. , ,Sure .

[0080] Furthermore, in step S103, based on the current selected position and optimal displacement of each segmentation point, each segmentation point in the next iteration process can be determined using formula (21), thus completing the update of the segmentation points, i.e.: ;(twenty one) At the same time, according to the preset convergence conditions described in the following formulas (22) and (23), including the function value convergence condition and the function value gradient convergence condition, it is determined whether the function value of the current objective function has converged.

[0081] The convergence condition for the function value is: ;(twenty two) The convergence condition for the gradient of a function is: ;(twenty three) in: ; If the current objective function value does not meet the above-mentioned preset convergence conditions, then repeat steps S101 to S103 to perform the next iteration until the objective function value meets the above-mentioned function value convergence conditions and function value gradient convergence conditions. This determines the optimized position of each segmentation point in the current iteration, thus identifying each optimal segmentation point. .

[0082] Figure 4 (a) in this application is a variable cross-section saddle-shaped weld groove provided in an embodiment of this application. x - zFig. 2 is a schematic diagram of a segmentation result of point cloud data on a plane, (b) is another variable cross-section saddle-shaped weld groove in x - z Fig. 3 is a schematic diagram of a segmentation result of point cloud data on a plane. As Figure 4 It can be seen that the contour shape size of the variable cross-section saddle-shaped weld groove also changes with the position, and the method of the embodiment of the application can achieve accurate segmentation of point cloud data of different saddle-shaped weld groove contour shapes.

[0083] Further, according to the optimal segmentation point The segmentation point set of each frame of laser scanning point cloud of the variable cross-section saddle-shaped weld groove three-dimensional point cloud can be determined as: ;(24) Finally, through the spatial mapping relationship between each frame of laser scanning point cloud, the variable cross-section saddle-shaped weld groove three-dimensional point cloud data can be accurately segmented into four segmentation segments by the optimal segmentation point, and can be respectively represented as: ;(25) The method of the embodiment of the application uses a composite objective function with length penalty to minimize the objective function, searches for the optimal segmentation point position by performing three-dimensional point cloud segmentation point change detection, minimizes the fitting deviation, and combines the maximum number of change points to realize fast optimization and positioning of the saddle-shaped weld groove point cloud segmentation point, and effectively improves the accuracy of the saddle-shaped weld groove point cloud data segmentation.

[0084] The thick plate saddle-shaped weld groove point cloud segmentation device provided by the application will be described below. The thick plate saddle-shaped weld groove point cloud segmentation device described below can be correspondingly referred to the thick plate saddle-shaped weld groove point cloud segmentation method described above.

[0085] Figure 5 Fig. 4 is a structural schematic diagram of the thick plate saddle-shaped weld groove point cloud segmentation device provided by the embodiment of the application, as shown in Figure 5 includes: a dimension reduction module 10, configured to perform dimension reduction processing on the acquired three-dimensional point cloud data of the saddle-shaped weld groove to obtain target plane point cloud data; a first processing module 20, configured to perform position numbering on each point in the target plane point cloud data, and determine a plurality of segmentation points from the numbered target plane point cloud data to determine a plurality of segmentation segments; a second processing module 30, configured to iteratively optimize the selected position of each segmentation point until the optimal segmentation point is determined, with the minimization of the objective function as the target; the objective function is determined based on the endpoint fitting deviation and the slope consistency deviation of each segmentation segment; The segmentation module 40 is configured to segment the three-dimensional point cloud data according to the respective optimal segmentation points.

[0086] It can be understood that the detailed function implementation of each unit / module described above can refer to the description in the foregoing method embodiments, which will not be described herein.

[0087] It should be understood that the above device is used to execute the method in the above embodiments, and the corresponding program modules in the device have similar implementation principles and technical effects to those described in the above method. The working process of the device can refer to the corresponding process in the above method, which will not be described herein.

[0088] The thick plate saddle-shaped weld groove point cloud segmentation device provided by the embodiment of the application can obtain spatially intersecting multi-layer multi-pass variable cross-section saddle-shaped weld groove three-dimensional point cloud data through online laser vision frame-by-frame scanning, convert the point cloud data to a target plane for data dimension reduction analysis, divide the point cloud data into multiple segmentation sections according to the spatial variation curve characteristics contained in the saddle-shaped weld groove, perform linear fitting on the segmentation sections, and estimate the consistency of the slope and the deviation of the straight line fitted by the end points of the segmentation sections, to construct a composite objective function for multi-objective optimization solving. The device can overcome the identification problem of the irregular contour of the cross section of the saddle-shaped weld groove, greatly improve the point cloud segmentation accuracy for the saddle-shaped weld groove, and accurately extract the point cloud data of different parts of the saddle-shaped weld groove.

[0089] Based on the method in the above embodiments, the embodiment of the application provides an electronic device, as shown in the figure, which can include a processor (Processor) 610, a communication interface (Communications Interface) 620, a memory (Memory) 630, and a communication bus (Communications Bus) 640. The processor 610, the communication interface 620, and the memory 630 can communicate with each other through the communication bus 640. The processor 610 can call the logic instructions in the memory 630 to execute the method in the above embodiments. Figure 6

[0090] In addition, the logic instructions in the memory 630 described above can be implemented in the form of a software function unit and sold or used as an independent product. When used, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the embodiments of the application.

[0091] ​Based on the method in the above embodiments, the embodiments of the present application provide a computer readable storage medium, which stores a computer program. When the computer program is run on a processor, the processor executes the method in the above embodiments.

[0092] Based on the method in the above embodiments, the embodiments of the present application provide a computer program product, which, when run on a processor, causes the processor to execute the method in the above embodiments.

[0093] It can be understood that the processor in the embodiments of the present application can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor can be a microprocessor or any conventional processor.

[0094] The method steps in the embodiments of the present application can be implemented in the form of hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in a random access memory (RAM), a flash memory, a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically EPROM (EEPROM), a register, a hard disk, a mobile hard disk, a CD-ROM or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor, so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and the storage medium can be located in an ASIC.

[0095] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in or transmitted by a computer readable storage medium. The computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through a wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) manner. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media sets. The available media can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0096] It can be understood that various numerical numbers involved in the embodiments of the present application are only distinguished for convenience of description, and are not used to limit the scope of the embodiments of the present application.

[0097] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for thick plate saddle weld groove point cloud segmentation, characterized in that, The method comprises: dimensionality reduction processing is performed on the obtained three-dimensional point cloud data of the saddle-shaped weld groove to obtain target plane point cloud data; each point in the target plane point cloud data is numbered in position, and a plurality of segmentation points are determined from the numbered target plane point cloud data to determine a plurality of segmentation sections; iterative optimization is performed on the selected position of each segmentation point with the objective of minimizing the target function until optimal segmentation points are determined; the target function is determined based on the endpoint fitting deviation and the slope consistency deviation of each segmentation section; the three-dimensional point cloud data is segmented according to the optimal segmentation points.

2. The thick plate girth weld bevel point cloud segmentation method of claim 1, wherein, The iterative optimization of the selected position of each segmentation point with the objective of minimizing the target function until the optimal segmentation points are determined comprises: Step S101: under a preset position constraint condition, the selected position of each segmentation point in the current iteration process is adjusted according to a preset displacement strategy, and the function value of the target function corresponding to the selected position of each segmentation point after adjustment is calculated; Step S102: the optimal displacement amount corresponding to the minimum target function value in each function value is determined; Step S103: each segmentation point in the next iteration process is determined according to the selected position of each segmentation point in the current iteration process and the optimal displacement amount; Steps S101 to S103 are repeated until the function value of the target function meets a preset convergence condition, and the optimal segmentation points are determined.

3. The thick plate girth weld bevel point cloud segmentation method of claim 2, wherein, The preset displacement strategy is determined based on equal division of a step function in the current iteration process; the step function is determined based on the total number of points in the target plane point cloud data and a decay factor in the current iteration process; the decay factor is determined according to the minimum target function value in the last iteration process and the initial function value of the target function.

4. The thick plate saddle weld bevel point cloud segmentation method according to any one of claims 1-3, characterized in that, Before the iterative optimization of the selected position of each segmentation point with the objective of minimizing the target function until the optimal segmentation points are determined, the method further comprises: determining the endpoint fitting deviation and the slope consistency deviation of each segmentation section based on the coordinate information of each segmentation point and the coordinate information of each point in each segmentation section; determining the target function based on the endpoint fitting deviation and the slope consistency deviation of each segmentation section and a preset segment length penalty term parameter.

5. The thick plate girth weld bevel point cloud segmentation method of claim 4, wherein, The determination of the endpoint fitting deviation of each segmentation section specifically comprises: determining a linear fitting function of each segmentation section based on the coordinate information of each segmentation point; determining the amplitude and the number of points in each segmentation section; determining the endpoint fitting deviation of each segmentation section based on the coordinate information of each point in each segmentation section, the linear fitting function, the amplitude and the number of points in each segmentation section.

6. The thick plate girth weld bevel point cloud segmentation method of claim 4, wherein, The determination of the slope consistency deviation of each segmentation section specifically comprises: determining the endpoint fitting slope of each segmentation section based on the coordinate information of each segmentation point; determining the least square slope of each segmentation section based on the coordinate information of each point in each segmentation section; A slope consistency deviation of each of the segmentation segments is determined based on a slope fitted at endpoints of each of the segmentation segments and a least square slope.

7. The thick plate girth weld bevel point cloud segmentation method of any of claims 1-3, wherein, The three-dimensional point cloud data of the saddle-shaped weld groove is dimensionally reduced to obtain target plane point cloud data, including: The three-dimensional point cloud data is subjected to point cloud coordinate decentralization processing to obtain decentralized three-dimensional point cloud data; The target rotation matrix is determined according to the decentralized three-dimensional point cloud data; The three-dimensional point cloud data is projected onto the target plane according to the target rotation matrix to obtain the target plane point cloud data.

8. A thick plate saddle weld groove point cloud segmentation apparatus, characterized by, It includes: a dimension reduction module for dimensionally reducing the three-dimensional point cloud data of the acquired saddle-shaped weld groove to obtain target plane point cloud data; a first processing module for numbering the positions of each point in the target plane point cloud data and determining a plurality of segmentation points from the numbered target plane point cloud data to determine a plurality of segmentation segments; a second processing module for iteratively optimizing the selected positions of each segmentation point to determine optimal segmentation points, with the objective of minimizing an objective function; the objective function is determined based on the fitting deviation at the endpoints of each segmentation segment and the slope consistency deviation; a segmentation module for segmenting the three-dimensional point cloud data according to each of the optimal segmentation points.

9. An electronic device, comprising: It includes: at least one memory for storing a computer program; at least one processor for executing the program stored in the memory, when the program stored in the memory is executed, the processor is used to execute the method as claimed in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. When the computer program runs on the processor, the processor executes the method as claimed in any one of claims 1-7.

Images