Method for recognizing geometric shapes of basic pixel point cloud surfaces and method for recognizing features of basic pixel point cloud surfaces

Abstract

The invention provides a method for recognizing geometric shapes of basic pixel point cloud surfaces and a method for recognizing features of the basic pixel point cloud surfaces. The method for recognizing geometric shapes comprises the following steps of: grouping basic pixels by utilizing feature significance of Gaussian images on the basis of a Gaussian mapping method, so as to obtain a plane-cylindrical surface-conical surface group and a spherical surface-toroidal surface group; recognizing planes, cylindrical surfaces and conical surfaces in the plane-cylindrical surface-conical surfacegroup through Gaussian image feature analysis; and constructing Laplace-Beltrami operators, calculating operator values of the Laplace-Beltrami operators, and recognizing spherical surfaces and toroidal surfaces in the spherical surface-toroidal surface group by utilizing a mean value and a variance of the operator values. The method for recognizing features comprises the following steps of: recognizing a geometric shape of a basic pixel point cloud surface; and extracting geometric parameters of the shape according to the recognized geometric shape of the basic pixel point cloud surface. Thetwo methods have the beneficial effects of being simple and feasible and good in pertinence, and being capable of strengthening the guidance quality of parameter extraction.

Description

technical field [0001] The invention relates to the field of computers, and is used for surface reconstruction technology. More specifically, it relates to a geometric shape recognition method of a point cloud surface of a basic primitive and a feature recognition method of a point cloud surface of a basic primitive. Background technique [0002] The surface of the physical sample is composed of multiple curved surface patches. After the point cloud model is segmented, multiple corresponding curved surface patch regions are formed. Since the region does not have corresponding geometric information, it is impossible to accurately reconstruct each curved surface patch. Analyze and study the point cloud surface Feature recognition is of great significance. The traditional point cloud reconstruction method does not carry out surface patch recognition, but uses methods such as implicit functions to uniformly model the point cloud model to obtain an overall approximate surface, or...

AI Technical Summary

Problems solved by technology

Existing surface feature recognition methods do not clearly distinguish the two processes

Images