Reproduce “Robust Estimation of Surface Curvature Information from Point Cloud Data” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jared Spang
Jared Spang from Columbia University develops the Voronoi Weingarten Map Estimator (VWME), a hybrid approach for robustly estimating surface curvature from noisy point cloud data. This method integrates the reliable normal estimation of the Voronoi Covariance Measure (VCM) with the Weingarten Map Estimator (WME), demonstrating improved resilience to significant noise compared to existing methods.
Key insights
- PCA-based normal estimation frequently suffers from a "flipping problem" where estimated normals point in the opposite direction, whereas VCM largely maintains correct orientation.
- K-nearest neighbor (K-NN) based neighborhood graphs are more robust for practical applications with noisy data compared to epsilon-ball (ε-ball) graphs, which can lead to disconnected points.
- The optimal choice of neighborhood size for curvature estimation varies significantly with noise levels and can introduce additional instability if not properly tuned.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.