Robust and accurate normal estimation in 3D point clouds via entropy-based local plane voting

Abstract

Abstract This paper introduces a robust and accurate normal estimation method for 3D point clouds. Our proposed technique is also robust towards noise and is capable of preserving sharp features in the input model. Our method presupposes that the normal of a point can be constituted from at least one in the normal set from the planes in its neighborhood, and its core idea is based on a local plane voting strategy, where each vote takes the entropy value and plane credibility into consideration. In addition, the average fitting residuals and the plane density are designed to further effectively deal with noise and non-uniformly sampled point clouds. The validity and reliability of our approach are confirmed by contrast to the most relevant state-of-the-art methods and by comprehensive experiments on synthetic and real-world data.