A Quad Tree Based Self-collision Detection Method for Cloth Simulation

Abstract

Regarding the self-collision detection efficiency during cloth simulation, this paper presents a quad tree based self-collision detection method. In this paper, we construct a quad bounding box tree for cloth according to the spatial location of its geometric primitives from top to down. And it divides the self-collision detection process into two stages. During the first rough detection stage, with normal cone method and detection of distance between triangle pairs’ centroids, we reduce triangles pairs for accurate elements intersection tests; during the accurate intersection stage, we filter out point-triangle pairs which impossible intersect using ipsilateral determination criteria rapidly. The final experimental result shows that the above optimization may greatly reduce the calculation workload of intersection detection and effectively improve the real-time while ensuring high verisimilitude.