Research On 3D Point Cloud Registration Algorithm

In the field of reverse engineering,digitalization of cultural relics and computer vision,the influence of measuring equipment,shape of objects and environment,resulting in the need to obtain a complete point cloud data model must be measured from multiple different perspectives.Point cloud registration combines point clouds under different coordinate systems into a unified coordinate system,which is an important step in obtaining the complete data model of the measured object.For a large number of point cloud data,the existing registration algorithm is either too long or the registration accuracy is too low,and it is difficult to achieve a balance between the two.To solve this problem,this paper proposes a point cloud rough registration algorithm based on feature extraction and matching and a normal distribution transformation algorithm based on feature point extraction.The specific work of this article are as follows:First of all,aiming at the problem that the existing rough registration algorithm is slow or low accuracy when dealing with a large amount of point cloud data,this paper proposes a point cloud rough registration algorithm based on feature extraction and matching.The algorithm according to the average curvature of each point to determine the retention point,and the retention point by the bump point to extract feature points.By extracting feature points from two cloud points to be registered respectively,the speed of the registration algorithm is improved.In the registration process,the corresponding point pairs are searched by FPFH feature and Hausdorff distance,and the RANSAC algorithm is used to eliminate the error point pairs so as to improve the accuracy of the correspondence relationship.Then,in order to deal with the problem that the existing normal distribution transformation algorithm deals with a lot of point cloud data,this paper proposes a normal distribution transformation algorithm based on feature points.The algorithm uses the feature point set of the point cloud to register,reduces the number of points participating in the registration,and removes the influence of redundant points.In the process of registration,the mixed probability density function is introduced,and the optimaltransformation is iteratively solved by the Newton algorithm to ensure the accuracy of the algorithm's registration.Finally,the validity of the proposed registration algorithm is verified both from the running time of the algorithm and the registration error,and the experimental results of the registration algorithm are demonstrated.Experimental results show that the proposed algorithm has obvious advantages in both accuracy and speed of registration when dealing with a large amount of point cloud data.