Morphology-constrained point cloud data registration sequence determination method

Abstract

The invention provides a morphology-constrained point cloud data registration sequence determination method in order to improve the registration efficiency of large-data-volume multi-view point cloud data, and belongs to the field of product reverse engineering. The local topography flatness of the point cloud is measured through principal component analysis, the point cloud in a flat area is segmented, a point closest to the centroid is extracted as a core point for adaptive simplification, the topography complexity of the multi-view point cloud is quantified through a deviation mean value from all sampling points in the point cloud to a fitting plane, and a registration sequence is determined through the topography complexity. The method is suitable for ordered multi-view point cloud data with overlapped areas, accumulated errors in the registration process can be effectively reduced on the premise that the point cloud scale is remarkably reduced, and the overall registration efficiency is improved.

Description

technical field [0001] The invention provides a shape-constrained point cloud data registration sequence determination method, which belongs to the technical field of product reverse engineering. Background technique [0002] In the field of reverse engineering, the mainstream raster projection equipment and binocular vision measurement equipment in the 3D scanning equipment are limited by the scanning angle of view during the scanning process, and it is usually difficult to obtain a complete data model at one time. The point cloud registration method unifies the scan data of multiple different views, that is, the multi-view point cloud data, into the same coordinate system. [0003] Retrieval of prior art documents found that in the paper "Object modeling by registration of multiplerange images" published in the academic journal "Image and Vision Computing" 1992.10(3):145-155, Chen et al. The coordinate system is determined as the reference coordinate system, and then new ...

AI Technical Summary

Problems solved by technology

[0004] To sum up, the existing point cloud data registration method mainly performs registration by sequentially adding point cloud data. When the sample point scale of point cloud data is large or the overlapping area of ​​an adjacent viewing angle is small, the registration efficiency is still not high. And it will form a large cumulative error, so how to realize the fast and accurate registration of large data volume and multi-viewpoint cloud is the research focus and difficulty in the field of reverse engineering

Images