Research And Application Of Point Cloud Registration Algorithm In Workpiece Pose Estimation

Point cloud pose estimation is based on the point cloud information of the target object to obtain its position and attitude in three dimensions.In recent years,scholars at home and abroad have carried out a lot of research work,but the accuracy and efficiency of three-dimensional point cloud processing under field conditions in industrial environments still need to be further improved.In this paper,we carry out relevant research work on the problems in point cloud preprocessing,point cloud segmentation and point cloud posture estimation in the process of robot intelligent picking of workpieces.The main research contents are as follows:(1)Spatial construction of point cloud is carried out by using KD-Tree algorithm to obtain spatial information of point cloud and realize K neighborhood query;The method of principal component analysis is used to estimate the normal vector of point cloud and obtain the curvature information.In view of high density point cloud,this paper adopts the method of direct filtering and voxel down-sampling to realize point cloud simplification.This method not only reduces the number of point clouds,but also retains some details and effectively removes noise points.(2)Aiming at the problem that the existing Fuzzy C-means(FCM)clustering algorithm is prone to fall into local optimum,this paper proposes Particle Swarm optimization(PSO)and FCM clustering method.In this method,the clustering center is searched by PSO algorithm and the fitness function is defined by FCM.The method proposed in this paper can achieve high precision clustering segmentation of parts point cloud.In addition,the experimental results show that the clustering effect of the proposed algorithm is better.(3)Aiming at the problem that traditional FPFH Feature descriptors are not effective in identifying symmetric parts point clouds in laboratory,an improved FPFH Global Feature Histograms(IGFH)is proposed in this paper.Firstly,an improved global feature descriptor IGFH is constructed by adding an orthogonal vector viewpoint component with viewpoint direction to the FPFH feature descriptor.Then feature descriptors and ICP precision registration were used to estimate the pose of the point cloud.Experimental results show that the improved algorithm proposed in this paper performs better in registration efficiency and accuracy.