Umbrella Curvature: A New Curvature Estimation Method for Point Clouds

This paper presents a novel method for surface curvature estimation. By using the k-nearest neighbors algorithm, an 8 neighbors ring is selected from point cloud for each point. Surface normal vector is estimated by neighbors coordinates. Local surfaces are classified using the point of interest and its 8 neighbors ring coordinates. In this paper surface curvature is calculated based on normal vector and neighbors ring coordinate. Surface curvature calculated using the novel method is called umbrella curvature. To evaluate how this method performs, the curvature of a sample surface was estimated by the umbrella curvature and surface variance method. The results show that the umbrella curvature performs well and is superior in terms of accuracy. However the method in computation is more expensive than variance method.