Continuous Collision Detection For Fabrics Based On The Hierarchical Culling

The flexible fabrics show the dynamic behaviors in 3D simulation typically,such as hysteresis,damping and anisotropy,but there is no usable accurate model that can totally simulate the complex deformation of the fabrics.In deformation simulation for fabrics,collision processing is a key technique.Effective detection and real-time response ensure the validity of the fabric simulation.Collision detection is divided into continuous collision detection and discrete collision detection.Compared with the discrete collision detection method which works on fixed time points,continuous collision detection works on time intervals and avoids the collision loss effectively.It is more suitable for simulation of fabrics which have a complex deformation process.The traditional continuous collision detection is computation intensive.In terms of a pair of triangular figures with collision possibility,fifteen primitives are dealt with precise detection computation.The computational consumption is pretty expensive for the high precision models during the physical simulation.In the actual collision queries,there is no need to accurately detect so many basic primitives.In 'order to reduce the redundant computation in the continuous collision detection,this thesis proposes a hierarchical culling method.Firstly,we do a pretreatment on basic primitives to avoid redundant detections caused by mesh connectivity during the model initialization.We also construct extra bounding box and bounding box tree to accelerate the collision detection.Then,we insert the hierarchical culling operation before accurate detection on basic primitives.High-level culling is executed to basic primitives which have no possibility to collide based on extra bounding box and improved bounding box overlap judging method,and low-level culling is conducted to basic primitives which do not need precise roots according to the algebraic method.Finally,we get the reliable accurate collision information by estimating the validity of roots in distance functions for basic primitives which are solved after hierarchical culling operation.The experimental results show that the method proposed in this thesis can not only effectively detect the collision information,but also have a robust culling feature when intensive deformation occurs.