Method and system for global registration between 3D scans

Abstract

The invention provides a computer-implemented method and a computerized device for global registration between a first point cloud and a second point cloud obtained by a scanning device on the same spatial scene of two independent instances. The method includes the step: extracting a first set of discriminant line pairs from the first point cloud and extracting a second set of discriminant line pairs from the second point cloud, wherein the discriminant line pairs have a higher discriminant power than randomly selected line pairs. In some embodiments, a plurality of matching line pair groups between the two discriminating line pairs are then determined according to a threshold criterion related to the relationship between the lines, the geometry of the lines, and the location of the lines, and a compass angle criterion related to the compass error of the scanning device. The method further comprises the steps of finding the most reliable corresponding relation between the two point clouds through voting, and then calculating a global transformation matrix; and finally, aligning the two point clouds by using a global transformation matrix. Embodiments of the present invention provide accurate and effective registration, particularly for building construction applications.

Description

technical field [0001] The present invention relates to processing three-dimensional (3D) scans, and more particularly to a computerized method and system for global registration between 3D scans. Background technique [0002] For visualization and further data processing of point clouds from 3D scans of the same spatial scene, it is necessary to transform these single scans into a common coordinate system. This process of finding a spatial transformation to align the 3D scans is called "registration". For two or more 3D point clouds with a common subset of points, the goal of 3D registration is to compute a rigid transformation that aligns these point clouds, providing an estimate of the relative pose between them. [0003] There are many existing registration methods, but most of them have limitations in robustness and processing time. Due to instability of the compass data or other issues, traditional methods may produce misalignment between the two scanned 3D models. ...

AI Technical Summary

Problems solved by technology

Due to unstable compass data or other issues, traditional meth

Images