For the next STORM seminar, Thibault Lejemble will give a presentation titled : Differential properties estimation in 3D point clouds

Abstract

In the context of 3D numerical acquisition, a point cloud is a common discrete representation of surface that is obtained when scanning an object by photogrammetry or using Lidar systems for instance. Local surface characterization is usually an important step for processing a point cloud for shape matching, registration, classification, editing and many other geometry processing tasks. A standard approach is to estimate the differential properties of the unknown points-sampled surface.
In this presentation, we present a method that estimates curvatures using the algebraic sphere regression. We propose a more accurate principal curvatures estimator and prove that the mean curvature estimator theoretically converge to the actual mean curvature of the surface when the support size tends toward zero. We also present an ongoing work on stability analysis in order to prove the robustness of the approach. Finally, we show our current results with a numerical evaluation comparing several state-of-the-art methods.

This seminar will take place online on Friday 29/01 at 12:30pm.
(If you want to attend the seminar, do not hesitate to ask for the zoom link.)