Multivariate analysis of curvature estimators

Abstract

The photogrammetry procedure is introduced as a cheap and -effective alternative to 3D scanning for constructing 3D point cloud data for reverse engineering applications. The data is then manipulated to generate CAD-ready models for FEA simulations or 3D printing purposes. A practical demonstration using model cars is conducted to show the effectiveness of the procedure in generating point cloud data, and various improvements are introduced. In particular, and innovative to this research work, a feature segmentation algorithm based on color information captured from photographs is implemented. This program, embedded within the photogrammetry process, allows for automatic selection of parts and features from an assembly model, providing a leading advantage over current 3D scanners, which capture spatial data only. Moreover, an intuitive method of enhancing the selection of required features of a triangular mesh during the segmentation process is introduced. This is achieved by manually marking the boundaries on the part of the object to be selected prior to taking photos. This has been proved to be more effective where areas of the object parts have similar colors.

Mesh Statistics for Robust Curvature Estimation

“…Even though similar studies have been done previously [MSR07, KLN08, MFM10, VKB16], our experiment involves a much wider gamut of possible data types, as well as approaches not known at the time of the previous works.…”