Cocone curved surface reconstruction method for maintaining object surface sample point edge characteristics

Abstract

The invention provides a Cocone curved surface reconstruction method for maintaining object surface sample point edge characteristics. The method is characterized in that edge characteristic reconstruction is optimized by correcting a normal estimation result and adjusting the Cocone angle threshold value of edge characteristic sample points. Gain optimization is performed on a localized curved surface sample used by PCA normal estimation so that the normal estimation result is corrected. An adjusting formula for the Cocone angle threshold value of the edge characteristic sample points is established so that the size of the angle threshold value of the edge characteristic sample points is adaptive to sample point distribution of an edge characteristic region. In the angle threshold value adjusting process, identification of the edge characteristic sample points is realized on the basis of Gauss clustering of the localized curved surface sample of the target sample points, and a cross-region long and narrow surface patch generated by adjustment of the angle threshold value of the edge sample points can be rapidly filtered on the basis of the predetermined scale threshold. Compared with methods in the prior art, defects of the reconstruction result of recesses and holes in the edge characteristic region can be substantially reduced.

Description

technical field [0001] The invention provides a Cocone curved surface reconstruction method for maintaining the edge features of sample points on the surface of an object, and belongs to the field of reverse engineering of products. Background technique [0002] When reconstructing edge features, the main problem of the Delaunay mesh filtering algorithm is that the reconstructed edge curves are incomplete and the adjacent facets of the edges are missing, which leads to a serious lack of precision in the subsequent NC machining tool path planning based on the reconstruction results. [1] . In view of the above problems, the current mainstream solution is to first extract the edge feature samples from the physical surface samples, and then construct the edge feature curves, and then use the feature curves to create constraints, and use the typical Delaunay grid algorithm under these constraints , such as the Cocone algorithm and its derivative algorithms, to reconstruct the su...

AI Technical Summary

Problems solved by technology

Since the Cocone algorithm estimates the sample point normal direction and determines the global optimal angle threshold based on the ideal Voronoi lattice model, it can theoretically guarantee the correctness of the reconstruction results only when the distribution density of edge sample points reaches infinity, while in practice the structured light 3D Scanners and other equipment sample the surface of the object, and the distribution density of the sampling data is always finite, which makes the Voronoi normal estimation result inaccurate, and the global angle threshold is generally difficult to adapt to the reconstruction of edge features, which eventually leads to the Cocone algorithm in the edge Reconstruction results at features often consist of holes or indentations

Dey et al. noticed that the Voronoi normal estimation based on the global Voronoi map is not suitable for Cocone local surface reconstruction. In the article "Localized Cocone surface reconstruction" (Computers & Graphics, 2011, 35(3): 483-491), the sample point adjacent to the largest internal angle is used The normal direction of the triangle surface is used as the normal estimation at the sample point, which in fact exacerbates the problem of inaccurate normal estimation, leading to further deterioration of the edge reconstruction results

Voronoi pole normal estimation depends on the Voronoi lattice shape of the target sample point, which is easily affected by the sampling data far away from the target sample point, and the estimation result cannot reliably reflect the distribution properties of the local samples on the surface at the target sample point

PCA (Princepal Component Analysis, Principal Component Analysis) algorithm can be used to estimate the normal direction of sample points, and it is more reliable to reflect the distribution properties of local samples. It can replace the normal direction of Voronoi poles in Cocone reconstruction. However, PCA normal direction estimation usually uses Euclidean neighbor point set As a local sample of a surface, it has poor adaptability to non-uniform point sets, and its estimation results may also be inaccurate, which exacerbates the problems of holes and edge dents in the Cocone reconstruction results.

[0004] In summary, the current Cocone surface reconstruction method has problems such as inaccurate normal estimation and difficulty in adapting the edge feature reconstruction to the global angle threshold. The method has become a technical problem to be solved urgently by those skilled in the art

Images