For the problem that in complex environment the low efficiency of object bounding box selection and updating directly lead to the extended collision detection time, a new hybrid CD algorithm relying on the tightness ratio of bounding volume and multi-layer modeling structure is proposed. The algorithm firstly divides object shape into four kinds by computing and comparing the tightness ratio of bounding box, using sphere bounding box, axis-aligned bounding box, oriented bounding box and ellipsoid bounding box respectively and carrying on intersection test in two bounding box for eliminating disjoint objects quickly and accurately. Meanwhile, in order to reduce the update time for constructing bounding box tree,the optimization method of multi-layer modeling structure which includes triangle,rigid body and model is proposed to construct bounding box tree.Finally,the experimental results show that,compared with using a single bounding box and the traditional hybrid bounding box collision detection algorithm,the proposed algorithm shorten the time of intersection test greatly and improve the efficiency of collision detection.