Automatic prismatic feature segmentation of scanning-derived meshes utilising mean curvature histograms

Abstract

This paper presents an enhanced curvature histogram based method to automatically segment scanning-derived triangle mesh models of physical prismatic mechanical parts into elemental feature patches. The segmentation task is in general very laborious due to the presence of noise in the scanned data. The scanning noise results in unsmooth mesh surfaces and more importantly, chamfered edges. The method consists of two successive parts in order to robustly deal with such issues caused by the noise. The first part extracts the sharpest features on the input mesh via analysing the histogram of mean curvature values at all the mesh vertices. The second part smooths the mean curvatures and non-sharp features are then extracted from the smoothed curvature histogram. The proposed two-part algorithm has been validated through the successful segmentation of elemental features on various scanned prismatic mechanical parts with no user-specified parameter tuning.

Keywords:

• feature segmentation
• prismatic mechanical part
• mesh model
• scanned data
• mean curvature

Additional information

Funding

Funding This work has been funded in part by the Natural Sciences and Engineering Research Council of Canada (NSERC).