New Bounding Volume Approach for Detecting Collision in Computer Animation

Abstract

This paper presents a method for fast-approximate collision detection between 3D models S undergoing rigid body motion known as oriented convex polyhedra R(S). By enclosing 3D models tightly, the fineness of detected collision can be enhanced. It is known that the large number of void areas which belongs to any 3D bounding volumes B(S) can affect the accuracy of collision detection system. Therefore, a way to compute R(S) using intersection of a set of halfspaces is described. The directions of these halfspaces are generated from calculating covariance matrix. To develop the tightest R(S), the quality of abutting corners by implementing Tribox Bounds method is improved. To detect collision between R(S), a straightforward approach by simply checking its interval pairs in local space system is performed. The proposed approach was implemented and a number of comparisons in terms of time and recorded collision with other B(S) were performed. From the conducted tests, R(S) performs well and might be a possible choice for detecting collision of 3D models undergoing rigid body motion.