Point cloud accurate registration method based on robust constraint least square algorithm

Abstract

The invention provides a point cloud accurate registration method based on a robust constraint least square algorithm. The method comprises the following steps: firstly, collecting point cloud data with a certain overlapping degree observed by a measured object from different visual angles according to a ground three-dimensional laser scanner, and carrying out coarse registration on a point cloudto be registered and a target point cloud through a PCA algorithm, so that the two point clouds have a relatively good initial position; secondly, establishing a kd-tree in the target point cloud, searching for points corresponding to the point cloud to be registered, and then reserving the points with the large feature degree according to the normal vector features of the point cloud; and finally, unifying the point cloud to be registered and the target point cloud to the same coordinate system through rotation and translation by using a robust constraint least square algorithm according to the obtained corresponding point pairs. The method effectively solves the problems that the original point cloud data size is large, the precision is not high and trigonometric function solving is complex, avoids falling into a local optimal solution in the registration process, shortens the registration time, and improves the precision of point cloud registration.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular relates to a point cloud precise registration method based on a Robust Constrained Least Squares (RCLS) algorithm. Background technique [0002] 3D point cloud registration is one of the key research issues in the field of computer vision, and has important applications in reverse engineering, SLAM, image processing, and pattern recognition. The purpose of point cloud registration is to solve the transformation matrix of point clouds with different attitudes under the same coordinates, and use this matrix to achieve accurate registration of multi-view scanning point clouds, and finally obtain a complete 3D digital model. [0003] For 3D point cloud registration, domestic and foreign scholars have conducted a lot of research, but most of the research methods tend to focus on the corresponding point pair search in the early stage, and improve the corresponding point pair search by e...

AI Technical Summary

Problems solved by technology

[0003] For 3D point cloud registration, domestic and foreign scholars have conducted a lot of research, but most of the research methods tend to focus on the corresponding point pair search in the early stage, and improve the corresponding point pair search by extracting relevant features of the point cloud such as normal vector and curvature. Efficiency, and then find the rotation matrix R and translation vector T by analytical solutions such as the SVD method or the quaternion method, but these methods do not consider the gross errors contained in the point cloud data and the cumbersome calculation of trigonometric functions, which is limited to a certain extent The accuracy of point cloud registration

Images