CN-122008223-A - Robot layered polishing method and device, electronic equipment and medium

CN122008223ACN 122008223 ACN122008223 ACN 122008223ACN-122008223-A

Abstract

The application discloses a robot layered polishing method, a device, electronic equipment and a medium, and relates to the technical field of industrial automatic processing, wherein the method comprises the steps of acquiring a scanning point cloud and a target point cloud of a workpiece to be processed, and obtaining a global allowance field on the surface of the workpiece to be processed by resolving based on normal line consistency propagation and directed projection; the method comprises the steps of determining the total layering number by planning the self-adaptive cutting depth of each machining point according to the global allowance field and combining the local geometric characteristics of a workpiece, discretizing the global allowance field in space based on the total layering number and the self-adaptive cutting depth to generate a plurality of point cloud subsets corresponding to each machining level, processing each point cloud subset, extracting and reconstructing structured grinding area data, and grinding according to the structured grinding area data. The application improves the self-adaptability and the polishing accuracy of polishing layering under the condition of rugged surface of the workpiece to be processed.

Inventors

DING GUOPING
SU ZHIFU
YANG MINGHUI
SONG CHUNSHENG
CHEN XI
LIU WENCHANG

Assignees

武汉理工大学

Dates

Publication Date: 20260512
Application Date: 20260319

Claims (10)

1. A method of robotic layered sanding, the method comprising: Acquiring a scanning point cloud and a target point cloud of a workpiece to be processed, and calculating to obtain a global allowance field on the surface of the workpiece to be processed based on normal line consistency propagation and directed projection; According to the global allowance field and combining with local geometric characteristics of the workpiece, planning the self-adaptive cutting depth of each processing point to determine the total layering number; Discretizing the global margin field spatially based on the total number of layers and the adaptive depth of cut, generating a plurality of point cloud subsets corresponding to respective processing levels; and processing each point cloud subset, extracting and reconstructing structured polishing area data, and polishing according to the structured polishing area data.
2. The method of claim 1, wherein the step of obtaining the scan point cloud and the target point cloud of the workpiece to be processed, and calculating the global residual field of the surface of the workpiece to be processed based on normal line consistency propagation and directed projection, comprises: performing voxel grid downsampling on the scanning point cloud, searching a neighborhood point set based on a tree structure, and predicting an initial normal vector of each point through principal component analysis; Constructing Li Mantu connected with the cloud tangential plane of the adjacent point, performing traversal propagation by utilizing a minimum spanning tree strategy, taking the dot product of the maximum normal vector of the adjacent point as a criterion, and performing consistent propagation on the initial normal vector to obtain a propagated normal vector; uniformly redirecting all the propagated normal vectors to the outer side of the surface of the workpiece to be processed by combining with external viewpoint constraint to obtain redirected normal vectors; determining nearest neighbors of each scanning point in the target point cloud based on the redirected normal vector, and calculating the directed distance from the current scanning point to the tangent plane of the nearest neighbors along the normal vector direction; and filtering out points with the directional distances of negative values, and forming a global margin field by the residual forward directional distances.
3. The method of claim 1, wherein the step of planning the adaptive depth of cut for each machining point based on the global margin field in combination with the local geometry of the workpiece to determine the total number of layers comprises: Carrying out statistical analysis on the global margin field, and processing to obtain a global maximum margin after removing outlier noise points outside a confidence interval; Constructing an initial self-adaptive cutting depth field according to local geometric features of the surface of the workpiece, wherein the local geometric features comprise Gaussian curvature and residual gradient of the surface of the workpiece; Performing smooth optimization on the initial self-adaptive cutting depth field to obtain an optimized self-adaptive cutting depth field; And calculating the total layering number based on the global maximum margin and the reference depth of cut in the optimized self-adaptive depth of cut field.
4. The method of claim 3, wherein the step of spatially discretizing the global margin field based on the total number of layers and the adaptive depth of cut to generate a plurality of point cloud subsets corresponding to respective processing levels comprises: Setting an iteration index i according to the total layering number by taking the target point cloud as a zero potential energy reference plane; calculating the space mapping distance of the lower reference surface of the current ith layer according to the optimized self-adaptive cutting depth field, wherein the space mapping distance of the lower reference surface of the current ith layer is the cumulative sum of the self-adaptive cutting depths of the previous i-1 layer; Calculating the spatial mapping distance of an upper reference surface of a current ith layer, wherein the spatial mapping distance of the upper reference surface of the current ith layer is the cumulative sum of self-adaptive cutting depths of the previous ith layer; And traversing the scanning point cloud, screening out points with global residual values larger than the lower reference plane space mapping distance and smaller than or equal to the upper reference plane space mapping distance, and generating an initial point cloud subset of the ith layer.
5. The method of claim 1, wherein the step of processing each of the subset of point clouds, extracting and reconstructing structured grinding region data, and grinding according to the structured grinding region data comprises: sequentially performing spatial clustering and geometric splitting processing based on a reference plane on each point cloud subset to obtain an optimized polishing sub-region cluster; Performing dimension reduction reconstruction on each polishing sub-region cluster to generate structured polishing region data, wherein the structured polishing region data comprises structured boundary description and space pose information thereof; And generating a polishing path according to the structured boundary description and the space pose information thereof, and polishing the workpiece to be processed according to the polishing path.
6. The method of claim 5, wherein the step of sequentially performing spatial clustering and reference plane-based geometric splitting on each of the point cloud subsets to obtain an optimized polished sub-region cluster comprises: Constructing a spatial index structure based on an initial point cloud subset of a current level, dividing the spatial index structure by adopting a spatial clustering algorithm based on density, and merging points connected in density into the same cluster to obtain at least one initial cluster; three non-collinear points in the space are selected in advance to define a reference plane, and the normal vector and the distance constant of the current reference plane are calculated; For each initial cluster, calculating the directed distances of all points relative to a reference plane, and judging whether the points in the current cluster are distributed on two sides of the reference plane; If the current initial cluster is detected to be distributed on two sides, the current initial cluster is forcedly split into two independent polishing sub-area clusters according to the sign of the positive and negative sides of the point on the reference plane; And returning to the step of judging whether the points in the current cluster are distributed on two sides of the reference plane or not until all the processed clusters do not cross the reference plane, and obtaining the optimized polishing sub-area cluster.
7. The method of claim 5, wherein the step of dimension-reducing reconstructing each of the clusters of ground sub-regions to generate structured ground region data comprises: Calculating three-dimensional coordinate covariance matrixes of all points in the current polished subarea cluster, and carrying out eigenvalue decomposition on the three-dimensional coordinate covariance matrixes to obtain three eigenvectors, namely three main components; Taking a plane formed by the first two principal components as the best fit plane of the polished subarea cluster, taking the two principal components as local coordinate system base vectors of the current plane, and taking the centroid of the subarea cluster as the local coordinate system origin; Transforming all three-dimensional points in the current polished sub-region cluster to the best fitting plane through orthogonal projection to obtain a two-dimensional projection point set; Calculating the convex hull of the two-dimensional projection point set to obtain a minimum convex polygon, wherein the minimum convex polygon comprises all projection points; And determining structured polishing area data according to the minimum convex polygon, the local coordinate system base vector and the local coordinate system origin, wherein the structured boundary description comprises the minimum convex polygon, and the spatial pose information comprises the local coordinate system base vector and the local coordinate system origin.
8. The utility model provides a robot layering grinding device which characterized in that, robot layering grinding device includes: the point cloud acquisition module is used for acquiring scanning point cloud and target point cloud of the workpiece to be processed, and calculating to obtain a global allowance field on the surface of the workpiece to be processed based on normal line consistency propagation and directed projection; the layer number determining module is used for planning the self-adaptive cutting depth of each processing point according to the global allowance field and combining the local geometric characteristics of the workpiece so as to determine the total layering number; The space discretization module is used for discretizing the global margin field in space based on the total layering number and the self-adaptive cutting depth to generate a plurality of point cloud subsets corresponding to each processing level; and the region reconstruction module is used for processing each point cloud subset, extracting and reconstructing structured polishing region data, and polishing according to the structured polishing region data.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the computer program being configured to implement the steps of the robotic layered sanding method as set forth in any one of claims 1 to 7.
10. A non-transitory storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the robotic layered sanding method of any of claims 1 to 7.

Description

Robot layered polishing method and device, electronic equipment and medium Technical Field The application relates to the technical field of industrial automatic processing, in particular to a robot layered polishing method, a robot layered polishing device, electronic equipment and a medium. Background In the fields of aerospace, wind power blade and high-end die manufacturing, surface finish machining of carbon fiber composite materials and large castings is a key link for determining product performance. Such workpieces often have the characteristics of rugged surface, uneven allowance, anisotropy and the like due to the influence of a forming process. The existing robot material removal path planning method is mainly divided into two types, namely offline programming based on an ideal CAD model, and the method cannot adapt to deformation and errors of an actual workpiece caused by casting, clamping and the like, so that machining precision is insufficient. The other type is a self-adaptive planning method based on three-dimensional scanning point cloud, which generates a material removal path by acquiring the real appearance of the workpiece, thereby obviously improving the adaptability to the actual workpiece. The current technology mostly employs parallel slices or simple equidistant biasing strategies to generate hierarchical paths. The stiff layered logic ignores the difference of local geometric characteristics of workpieces, and is extremely easy to cause process accidents when processing workpieces which are sensitive to contact force, such as composite materials and the like, wherein when a allowance gradient is severely fluctuated or a large curvature area is adopted, a fixed cutting depth is forcibly adopted, so that a grinding head is excessively loaded instantly, a force control system is instable and even a material is damaged, and when the allowance is flat, a conservative cutting depth is set, the processing efficiency is low, and the requirement of fine grinding is difficult to meet. Therefore, how to improve the self-adaptability and polishing accuracy of polishing layering under the conditions of rugged surface and different thickness of the workpiece to be processed is a problem to be solved at present. Disclosure of Invention The application mainly aims to provide a robot layered polishing method, a device, electronic equipment and a medium, which aim to solve the technical problems of how to improve the self-adaptability and polishing accuracy of polishing layering under the conditions of rough surfaces and different thicknesses of workpieces to be processed. In order to achieve the above purpose, the present application provides a robot layered polishing method, which comprises: Acquiring a scanning point cloud and a target point cloud of a workpiece to be processed, and calculating to obtain a global allowance field on the surface of the workpiece to be processed based on normal line consistency propagation and directed projection; According to the global allowance field and combining with local geometric characteristics of the workpiece, planning the self-adaptive cutting depth of each processing point to determine the total layering number; Discretizing the global margin field spatially based on the total number of layers and the adaptive depth of cut, generating a plurality of point cloud subsets corresponding to respective processing levels; and processing each point cloud subset, extracting and reconstructing structured polishing area data, and polishing according to the structured polishing area data. In an embodiment, the step of obtaining the scanning point cloud and the target point cloud of the workpiece to be processed, and calculating the global residual field on the surface of the workpiece to be processed based on normal line consistency propagation and directed projection includes: performing voxel grid downsampling on the scanning point cloud, searching a neighborhood point set based on a tree structure, and predicting an initial normal vector of each point through principal component analysis; Constructing Li Mantu connected with the cloud tangential plane of the adjacent point, performing traversal propagation by utilizing a minimum spanning tree strategy, taking the dot product of the maximum normal vector of the adjacent point as a criterion, and performing consistent propagation on the initial normal vector to obtain a propagated normal vector; uniformly redirecting all the propagated normal vectors to the outer side of the surface of the workpiece to be processed by combining with external viewpoint constraint to obtain redirected normal vectors; determining nearest neighbors of each scanning point in the target point cloud based on the redirected normal vector, and calculating the directed distance from the current scanning point to the tangent plane of the nearest neighbors along the normal vector direction; and filtering out points with the directional dis