Research On 3D Point Cloud Registration Technology With Low Overlap

3D point cloud registration refers to that the source point cloud is aligned to the target point cloud through coordinate transformation based on the target point cloud,and the two can overlap to the maximum extent.It is widely used in the fields of 3D modeling,SLAM,reverse engineering and 3D scanning.In practical applications,the data itself has defects or may be blocked,which makes the overlap area between the two point clouds is very small and the registration challenge is greater.Therefore,the study of low overlap rate 3D point cloud registration has very important practical value.The main research work of this paper is as follows:(1)Aiming at the problems of high difficulty and low accuracy in the registration of two point clouds with low overlap rate,a point cloud registration method combining the clustering area block and convex optimization problem is proposed.First,use the curvature characteristics of the point cloud to establish a multi-scale descriptor to ensure the integrity of the point cloud data and minimize the redundant data;second,use the angle difference of the multi-scale descriptor to cluster and partition the correspondence relationship to obtain the the overlapping area;finally,the point clouds and their corresponding relations are substituted into the convex optimization problem,outliers are removed and the corresponding relations are optimized,the coarse registration is realized and the ICP algorithm is used for refinement.(2)The advantages and disadvantages of the above-mentioned low-overlap 3D point cloud registration methods are analyzed,and a 3D point cloud registration method based on the concave-convex characteristics is further proposed.Combining the advantages of the ISS and the Harris algorithm,establish a mathematical model to remove the points that are too flat and too bumpy on the surface of the point cloud,so that the seed points of the flat area are relatively uniform,and the non-flat area The seed points are relatively concentrated,which effectively reduces and refines the seed points,and then performs the clustering area partition based on the refined seed matching set,and uses the convex optimization problem to complete the registration.(3)Experiments are carried out in the compiled environment of MATLAB,and various models are used for experimental verification of the two methods,which proves the accuracy and high efficiency of the proposed method for low overlap rate 3D point cloud registration.