The Application Of Spherical Harmonic In 3D Model Retrieval

With the development of Internet and computer technology, the application domain of 3D model is also getting more and more widespread. Since constructing a highly lifelike 3d model is a very time-consuming and trouble thing, it becomes more and more important to reuse the existing 3D models. Facing the huge 3d model database, how to rapidly search the suitable model becomes another popular topic after the image and video retrieval.The basic steps of 3d model retrieval include model feature extraction, feature description and similarity comparison. This is key point of the present research. Among them, the feature extraction and the feature description of 3D model play the decisive role in the 3D model retrieval. Therefore, the model feature extraction and the feature description is the focus of this paper.This paper include four chapters. In the first chapter, we first present the application perspective and the significance of 3D model research engine study, then introduce each kind of descriptors in existing 3D model retrieval technique and some retrieval technique, Finally we give an example of typical 3D model search engine. In the second chapter, we first describe the necessity of normalization, then we introduce PCA , translation, and voxelization in detail. In the third chapter, we first introduce the basic concept of spherical harmonic. Then, we give the experiment to compare three spherical harmonic model retrieval method.In the first experiment, we introduce a spherical harmonic model retrieval based on beam min distance. The concrete method constructs a function f(u) on spherical surface S2 which used to describe 3d model shape information. Then by spherical harmonic transform, the energy of the coefficients of spherical harmonics forms the feature vector.In the second experiment, we introduce a spherical harmonic model retrieval based on voxelization. The basic idea is similar with the first test. The difference lies in the model normalization and retrieval region of 3D model.In view of insufficiency of the two experiment above mentioned, we propose a spherical harmonic model retrieval method based on beam maximum distance, and give certain revision to the sampling process. The basic idea is: we select the distance maximum R of the center of mass beam and the triangle through sheet point of intersection's, and calculate each angle between the apex and center of mass. The experiment indicated that this method raised the retrieval speed, expanded the retrieval region.We conclude the whole thesis in fourth Chapter, with a brief discussion of future research directions .