Local increasing type curved-surface reconstruction method of large-scale point cloud data

Abstract

In order to improve the 3D curve surface fitting precision under point cloud data, the invention provides a method for carrying out curved-surface increasing reconstruction of point cloud data in a priority sequence. The method comprises the following steps that: (1), a point normal vector in point cloud data is calculated roughly and a clustering analysis is carried out according to the normal vector to obtain a point cloud set reconstructed based on first-level priority; (2), for the point cloud set, points in the point cloud set are transverse and local curved-surface fitting is carried out on each point, and a local three-dimensional curve surface with actual geometry and topology is reconstructed; (3), according to the adjacent curve surface with first-level reconstruction, increasing type local fitting is carried out on point cloud data, not being listed into first-level reconstruction, of the point cloud set and a three-dimensional curve surface corresponding to the point cloud data is reconstructed; and the steps is repeated continuously until all point cloud data are reconstructed; and (4), a final integrated three-dimensional reconstruction curve surface model is obtained by using a local function global mixing method.

Description

Technical field [0001] The present invention relates to the technical field of 3D surface reconstruction of an object, and more specifically, to a method for constructing a three-dimensional model of large-scale point cloud data by using a local growth type surface reconstruction method. Background technique [0002] For a long time, how to accurately and efficiently obtain the geometric shape information of real-world objects has been a fundamental problem in large-scale projects to realize the digitization and intelligence of design, display, data archiving and error detection. Traditional 3D scanning equipment has significant defects such as narrow and fixed viewing angles, short effective measurement distances, and high requirements for the scanning environment. Its application is still limited to indoor geometric measurement of small objects at close range. This technology is difficult to apply to large-scale environments, and its harsh requirements are difficult to meet in ...

AI Technical Summary

Problems solved by technology

Noise can also make topological identification difficult

like figure 1 Difficulties in model reconstruction from point cloud data listed in the prior art: (a) sharp corners require high sampling rate, even if the noise is not strong; (b) when two surfaces are relatively close, noise leads to topological singularity; ( c) Highly bursty noise; (d) Wrong assignment of surface directions on both sides of the sharp corner

[0007] Existing methods in the state of the art usually only partially solve the above problems under the premise of some strong assumptions, such as sampling rate, continuity of surface or amount of noise

Images