3D Skeleton Based 3D Model Retraival

In recent years, technology in fields, such as 3D printing, magnetic resonance imaging, 3D scan, 3D rendering, 3D reconstruction technology, has enjoying a spurt of progress. 3D model has been widely applied to the medical, construction, industry, film, video game, the Internet, e-commerce and other fields, providing great convenience in medical diagnosis, industrial manufacturing, film special effects, animation, virtual reality and so on. Along with the wide application of 3D model, the 3D model data grows explosively. 3D model retrieval has become an urgent demand.The 3D model, which is usually defined by voxel point or surface mesh, suffering a huge amount of data, cannot be directly used for 3D model retrieval. The skeleton, a one dimensional description with a construction of graph or tree of the original model, which has many wonderful characteristics, such as small amount of data, topological homotopy with the original model and so on, is an ideal feature of 3D model. skeleton matching algorithm is used to measure the similarity of different skeletons, which reflects the similarity between models. In this paper, 3D skeleton extraction and skeleton matching are studied and combined together to build 3D skeleton based 3D model retrieval system.First of all, we study and analyze the 3D skeleton extraction algorithm, such as thinning, distance transform based algorithm, the general field based algorithm, surface shrinkage, describe the characteristics and the application scope. Based on the global importance measure based skeleton extraction algorithm, a new algorithm is proposed which can extract skeleton, thin, connected and containing few noise branch.Then, the skeleton matching algorithm is studied. Among the existing skeleton matching algorithm, the algorithm based on skeleton path similarity, which achieves high matching accuracy, is considered to be a outstanding work. Unfortunately this algorithm applies only to the 2D model. In this paper, through multi angle projection and skeleton endpoints sequence labeling method, this algorithm is extended to the case of 3D model.Finally, by combining skeleton extraction and skeleton matching together, skeleton based 3D model retrieval system is build. The algorithm in this paper achieves competitive results in the Experiment.