Surface reconstruction via cooperative evolutions

Real-world point clouds data acquired by depth cameras or laser range scanners are usually imperfect and contain noises, outliers, and missing regions, which makes the reconstruction remain challenging. In this paper, we present a novel evolution-based method for surface reconstruction from point clouds. Our method evolves two deformable models from the interior and exterior of the input points respectively, i.e., one model expands to the points from its interior and the other model shrinks to the points from its exterior. Both deformable models evolve simultaneously in a cooperative and iterative manner. The evolution of each model is driven by an unsigned distance field as well as the other model and terminates when the two models are close enough with each other. A central surface is then extracted as the final reconstructed surface. Normal estimation is widely used in surface reconstructed algorithms, but in our method, we only take advantage of direction estimation without orientation. Experimental and comparison results have shown the feasibility and robustness of our method and its significant advantages in many specific situations, such as point clouds with thickness, irregular holes, or noises. (C) 2020 Elsevier B.V. All rights reserved.