Parallel Construction for Model-Adaptive Convex Bounding Polyhedron

TANG Lei,SHI KanLe,YONG JunHai,GU HeJin,XU Xin

DOI: https://doi.org/10.1360/n112014-00019

2014-01-01

Abstract:Bounding box is extensively used to accelerate algorithms such as geometric intersection, ray tracing and collision detection in computer graphics and computational geometry. Convex bounding polyhedron, which is a generalization of bounding box, can approximate a point set with higher tightness compared to the corresponding bounding box in usual cases. This paper proposes a method to construct the convex bounding polyhedron from a point set. Based on the coarsely approximate convex hull, the algorithm generates k normals by k -means algorithm. The algorithm searches the tangent point to generate a cutting plane along each normal on GPU and then constructs the convex bounding polyhedron by intersection of the cutting planes. Experiments illustrate that the algorithm can construct tighter convex bounding polyhedron efficiently and this advantage accelerates collision detection algorithms.