Fast Feature Curve Extraction for Similarity Estimation of 3D Meshes

Abstract

Feature extraction from 3D triangle meshes is a very popular and important task that could contribute to many scientific fields such as computer vision, pattern recognition, medical 3D modeling, etc. However, the main challenge is not just finding corners and edges of 3D models but to automatically extract connected clusters of vertices that jointly represent a feature curve. This paper presents an approach for feature curve extraction and similarity evaluation among feature curves of the same or other models robust to differences in scale, resolution quality, pose, or partial observation. The proposed approach could be used, as a pre-processing step, in many other applications like registration, partial matching, tracking, object recognition, etc. Extensive evaluation studies and experiments carried out using a variety of different models and use cases, verify that the proposed approach achieves accurate feature curve extraction and categorization, robust to several constraints like scale or resolution.