Collision Detection Of Morphing Bodies Based On Skeleton Animation

For applications of animations or games, detecting collisions among objects in avirtual scene is a classical problem in the fields of computer graphics and animations.Whether collision detections (CDs) can be executed both quickly and precisely or notdefines how comfortable and real the user experiences for a virtual scene can be, andthus has a direct affect upon the result of expression.Since objects in a3D scene are meshes composed of layers of triangle patches,whose number could easily reach a scale of thousands, if not tens of thousands, in asingle object, it leads to unaffordable time costs for CDs confined strictly with thetriangle meshes. Therefore, classical solutions against CD in3D scenes are to employvolumes,which comes with regular shapes, and are easy in terms of performing CD, tocover the triangle meshes in order to rule out obvious collision-free situations,reducing the computation costs resultantly. That is called the bounding-box (BB)method, and is proven able to decrease the time cost of CDs significantly.There are two kinds of rivial BBs. One of them is a rectangular box sharing samecoordinate ranges with the triangle mesh, while the other is the minimum round ballcovering every vertex in the mesh. In practice, rectangular BBs are more welcomeddue to the ease in computing the coordinates. In order to maximize the efficency of themethod and prevent mis-detections as many as possible, however, BBs are supposed tobe as firmly fitted against the object surface as possible, which leads to multiplevariations of rectangular BBs. Those variations are better in terms of tightness, whileare more complicated and involved to compute. In addition, these BBs are allcomposed from multiple flat planes, and are unable to achieve tightness against curlysurfaced objects by definition.As with the rapid development and evolvement of graphic softwares andhardwares, the objects in virtual scenes increasingly resemble the real ones. Obviouslyenough are the majority of objects consisting of some certain structure in the real world,rather than being simple collections of numerous vertices as the triangle meshes are in virtual scenes. That is to say, employing this pattern in modelling the virtual objectshelps producing animated scenes with great feel of reality, which is, in fact, the methodof skeleton animation (SA) that sees more and more popularity in the recent years.With SA, it’s far more easy and storage-saving to present motions of objects such ashuman, animals, and machines, compared with traditional vertex-based animations,which earns the method of SA a wide welcome in run-time rendering applications suchas computer games and computer simulations. When employing SA, on the other hand,the moving parts of the objects are usually rotating instead of translating in theanimation sequences, adverse to the usage of some kinds of traditional BBs. Inaddition, parts of human and animal bodies are usually more cylindrically orspherically shaped, meaning that traditional flat-surfaced BBs are unable to achieveideal tightness, thus decreasing the serviceabilities of the BB method. To overcome this,we present the cylindrical BBs, together with the CD algorithms among them, and putthem into use of CDs of morphing bodies. Test-run results show a significantimprovement of performance compared to doing the CD directly with the trianglemeshes, while winning the traditional OOB methods by some advantages.