Deformable Modeling And Rendering Of The Key Technologies To Achieve

As the fast development of computer hardware and software in the recent years, the computer graphics is playing a more and more important role in the fields of video games, movies, simulation system and so on. In most of these applications, the simulation of deformation is necessary. The methods used to simulate deformation can be classified to two types: non-physical method and physical-based method. As the performance of the 3D scanner improved highly, it's possible to gain very detailed models represented by point-cloud. Since the point-cloud models don't need any topology information, they can be easily operated without worrying about keeping of topology consistence. But the vertex number of point-based models is more and more than the mesh models. How to organize these data is very important. And because the point model is represented by completely discrete points, how to produce continuous surface without holes is another important problem to be solved.In my thesis, I focused on the meshless deformation based on shape matching method. This method was first proposed by Muller in 2005 and attracted many interest because it's simple to model, very fast and robust and easy to extend. This thesis introduces this shape matching method and two methods expanded from it in details and present another method expanded from the shape matching method. In our expanded method, we use KD-tree to organize model data, and use the k-nearest neighbors as the shape matching region. Our method could also do the fast summation operation according to the hierarchical of KD-tree and then deform the model using the shape matching method. Our method is very simple and easy to implement and stable unconditional.We also introduce the render technique of point-based model. Consistent with the deformation method, we employ a render method based on KD-tree, so that the deformation result could be used to render directly without other additional operations. In our method the data is organized with the balance KD-tree. Every node in the KD-tree stores some properties such as position, normal, normal cone and color used for rendering. Our method also improves the rendering efficiency through techniques such as View-Frustum Culling, Back-Face Culling and Level of Detail.At last, we employ the techniques introduced in the thesis to simulation destroy result of objects because of explosion. The method is also used to simulation cloth animation.