Point Cloud Registration Algorithm For Optimizing The Correspondence Of Paired Point Clouds

With the continuous development of 3D scanning technology,the application of 3D point cloud data in various industries has become more and more extensive.For example,3D point cloud data processing technology has been applied to the aerospace industry,automobile manufacturing,medical and health industries,etc.closely related to life.Industry.Due to the influence of various factors such as the shape and size of the object,the collection equipment,the on-site environment,etc.,in the actual point cloud data collection process,the scanner can only collect the point cloud data model of the object under a certain angle of view under the measuring station.The object point cloud model needs to transform the point cloud data collected from multiple measuring stations into the same coordinate system through space coordinates.Therefore,point cloud registration technology is a crucial part of 3D point cloud data processing.This article first denoises the acquired point cloud data,and then registers the point cloud after denoising.The main contents are as follows:(1)A denoising algorithm combining manifold and graph Laplacian regularization is proposed.First,the manifold patches are constructed by the point cloud data model.This article assumes that manifolds are embedded in the sampling,and the number of manifold patches is determined according to the shape and size of the point cloud;then the manifold patches are used to reduce their self-similarity.In the end,this paper constructs a discrete graph,uses the Tulaplace regularizer to approximate the size of the manifold defined in the continuous domain,and proposes a patch distance measurement method to quantify the similarity between plates.Experimental results show that the proposed algorithm can better retain the edge features and significant geometric features of the point cloud,and the denoising effect is better and the error is smaller.(2)A point cloud registration method that optimizes the correspondence between paired point clouds is proposed.First,use the fast feature histogram and feature optimization to generate the initial corresponding relationship;then,the initial corresponding point set is obtained by checking whether the corresponding points meet the nearest neighbor principle,and then the initial point set is judged by the L2 norm ratio,and the basic selection is The corresponding point set of the correct correspondence relationship is finally calculated by alternately optimizing the transformation matrix of the point cloud to realize the precise registration of the point cloud.The experimental results show that compared with the traditional algorithm,the algorithm in this paper has faster registration speed,higher registration accuracy,and has good robustness to point clouds with low overlap and high noise.