Research On Workpiece Recognition And Positioning Based On 3D Laser Scanning

This paper aims at the recongnition and positioning of specific workpieces in industrial production.It mainly studies the target recognition algorithm and positioning method based on three-dimensional laser scanning point cloud data.The main research contents are as follows:1)Analyzed the era significance of the topic of this paper,and introduced the development status of laser scanning technology and object recognition method based on point cloud,and led to the solution of object recognition and positioning based on laser point cloud.2)Based on the data characteristics of laser scanning depth maps,relevant researches on the extraction of adjacent points and image segmentation are conducted.The concept and search method of the conditional K-proximity point are proposed.And the bidirectional density asymmetry of the laser scanning point cloud is solved to extract the neighboring points,while using the depth The map with topological information accelerates the Euclidean segmentation process.3)The general idea of the recognition method in this paper is to extract the edge of the main plane of the workpiece first,and then complete the recognition and positioning on this basis.Edge extraction reduces the magnitude of processing data while retaining useful information for recongnition and positioning.This paper improves on the traditional method of random sample matching plane fitting,and proposes a more efficient RANSAC algorithm based on random point local normal.Combined with global least squares method,the main part of the workpiece is quickly and accurately implemented.Finally,the principal plane information and the edge extracted based on the scan line method are used to obtain the edge features of the main surface of the workpiece.And in this process,the occlusion relationship decision is made according to the nature of the edge.3)Finally,based on the extracted feature edges,this paper proposes a layered progressive recognition algorithm,and gives the initial pose for accurate ICP location in the recognition process.The progressive recognition algorithm consists of three layers with increasing complexity.The first layer is the basis of the second layer.Each layer algorithm has a large possibility of completing the recongnition task,and only starts the lower layer algorithm when the recongnition task is not completed.The first layer proposes an O(n)algorithm for computing the farthest point pair in the point cloud to express the size characteristics of the workpiece.The second layer is based on the first layer,with the farthest point pair as the reference,and features are constructed by the correspondence between the distance and the angle.The third layer recognizes the relative relationship of all pairs of points as features.The three methods are progressive in the complexity of the processing problem,and Adopting a hierarchical progressive recognition strategy makes the recognition more efficient.