Point-to-point model based universal 3D (three-dimensional) surface matching method

Abstract

The invention provides a point-to-point model based universal 3D (three-dimensional) surface matching method, which comprises the following steps: rapidly searching overlapped regions among point cloud data sets on a 3D surface, and marking point sets in the overlapped regions; carrying out data organization on point cloud data sets on a reference surface, and establishing topological relations among all node data in the point cloud data sets on the reference surface; according to a point-to-point model, selecting a specific identical point correspondence rule to carry out searching, and establishing a one-to-one identical point corresponding relation between a registration surface and the reference surface; establishing an error equation according to identical point pairs obtained through searching, and establishing a normal equation according to the established error equation and based on a principle of least squares, then calculating the normal equation so as to obtain an estimated value of the maximum likelihood registration parameter; and according to the calculated registration parameter, carrying out coordinate conversion on a to-be-registered surface. By using the method disclosed by the invention, a surface matching algorithm is unified theoretically, and an appropriate identical point correspondence rule can be selected according to the characteristics of to-be-registered data, therefore, the method has good expansibility.

Description

technical field [0001] The invention relates to the technical field of surveying and mapping, and more specifically, relates to a general least squares 3D surface matching method based on a point-to-point model. Background technique [0002] 3D surface matching (generally also called surface registration) is a frontier research issue in the field of computer vision and digital photogrammetry. Its main purpose is to establish the coordinate system conversion relationship between two three-dimensional The registration of the formed point cloud data point set is a key step to realize the seamless splicing of point cloud surfaces. The transformation relationship is generally described by all or some of the seven positive transformation parameters such as scale, rotation, and translation. There are two key factors in the implementation of registration: 1) choose an appropriate mathematical model to describe the geometric constraint relationship between 3D surfaces, and then accu...

AI Technical Summary

Problems solved by technology

[0007] To sum up, the existing mainstream 3D surface registration methods have not been able to form a unified and general adjustment mathematical model, and there is a lack of a general registration model in the prior art, which can theoretically organically carry out existing registration methods. Therefore, it is necessary to seek a general 3D surface registration model that can adapt to the definition rules of different points with the same name

Images