Improved Euclidean Clustering and Segmentation Algorithm for Workpiece Identification

Abstract

Robots with binocular vision often use template matching to identify the parts to be processed, which requires segmentation of the original point cloud data to obtain a separate target point cloud. The original point cloud has the characteristics of unstructured, dense and remote, and the traditional Euclidean clustering algorithm directly acting on the 3D point cloud has the problem of low segmentation accuracy, and the accurate segmentation of the point cloud target has always been a difficult problem in target detection. In this paper, an improved Euclidean clustering algorithm is proposed, under the constraint that the traditional Euclidean clustering segmentation only considers the Euclidean distance, the voxelized grid method is used to achieve downsampling of the point cloud data, and then the normal vector angle constraint is added, which can effectively improve the problem of low segmentation accuracy of the traditional Euclidean clustering algorithm, and also achieves good results in the actual production process after testing and verification of the public dataset.