3D Mesh Segmentation And Application In Retrieval

In recent years, with the rapid development of the three-dimensional scanning technology and computer graphics, as well as the increasing of three-dimensional models, three-dimensional mesh segmentation technology has become an active field of research and was widely used in many areas of computer graphics. However, the technology still has a lot of challenges: (1) There is no uniform criterion of three-dimensional mesh segmentation technology. In recent years, in order to imitate the human visual perception, most researchers pursue a meaningful segmentation result. However, human visual perception is complex; In addition, there are no uniform segmentation models, different segmentation algorithms select the respective models, which can not directly compare the segmentation results, making a meaningful segmentation result is very subjective. The attempt to imitate human visual perception to get meaningful segmentation is still a problem. (2) Semantic-based segmentation algorithm. Get the segmentation results based semantic, rather than be a part of a model (such as: legs, arms, etc.). (3) Find three-dimensional mesh segmentation technology applications, to solve specific application problems. Different segmentation algorithms are suitable for different applications, in accordance with the existing algorithms, to find the appropriate application is also a challenge.In this paper, we study around the three-dimensional mesh segmentation, primarily about the application areas of the three-dimensional segmentation algorithm, a new sub-parts retrieval method based an interactive segmentation algorithm is proposed. Existing retrieval technology also is aimed at a comparison between single models, while in model design, shape restoration applications often need to only a particular part of models. Moreover, searching for a specific part inside a 3D model is even more challenging. The method based statistics can get better search results, but does not support part-in-whole queries and partial matches; skeleton extraction and graph-based methods, although to support partial matches, have more stringent requirements about the structure of three-dimensional models . Thus, we present a method that finds analogies among parts of three-dimensional models by segmenting them and creating a signature for each part. To find analogies we first partition all objects in a database into parts by an interactive segmentation method. These parts of the models will be organized into a small database; next, we can retrieve the most similar parts from all models in the database based on shape distribution method and the PDF-L1 criterion. The experimental results show that we get better search results.