Point cloud registration algorithm based on topological characteristic

Abstract

The invention provides a point cloud registration algorithm based on a topological characteristic. Topological characteristic points (border characteristic points and highlight characteristic points) of the point cloud are extracted and are used for calculating an initial rotation matrix and an initial translation vector at the initial point cloud registration stage of the algorithm. A basic idea of an iterative closest point (ICP) algorithm is adopted at the accurate registration stage, and the method of choosing a registration element and determining a corresponding point set is improved. The registration element chooses the topological characteristic points, and the closest point of a neighbor center of gravity is chosen as the corresponding point when the corresponding point set is determined. The method introduces the topological characteristic, reduces the number of the extracted points and ensures matching effect because the topological characteristic contains abundant point cloud characteristics. The method takes the neighbor characteristic into consideration when determining the corresponding points, thereby enhancing robustness of the algorithm to noises.

Description

technical field [0001] The invention relates to computer vision technology, in particular to an algorithm for realizing three-dimensional point cloud registration based on topological features. Background technique [0002] Three-dimensional models are of great significance in the fields of virtual simulation training, cultural relics protection, and medicine. With the development of modern 3D scanning technology, the processing of point cloud models has become a research hotspot in recent years. An important aspect of point cloud processing is the registration of two point cloud data obtained from the same scene. [0003] According to the registration process, the registration of point clouds can be divided into: coarse registration and precise registration. Coarse registration is to calculate the registration transformation parameters by extracting the corresponding control points of different viewing angle data sets. However, due to the inability to obtain accurate cor...

AI Technical Summary

Problems solved by technology

The ICP algorithm can obtain relatively accurate registration results and is widely used, but it also has many shortcomings: the algorithm assumes that one point set is a subset of the other point set, which is difficult to meet in many cases; the algorithm is searching for corresponding points. In the process, the calculation cost is very high; in the basic ICP algorithm, when looking for the closest point, the algorithm searches for the closest point of the Euclidean distance. This assumption is relatively arbitrary, and it will generate wrong corresponding points

Images