Study And Implement On Spherical Harmonc Based 3D Models Retrieval System

The 3D-models have become more widely available with the development of the computer hardware and software, 3D-models acquisition. The 3D-models quantity is explosive growth, and more and more 3D-models database appeared. Reuse a 3D model is an important thing, because constructing a 3D-model is complex. To fully utilize the existed 3D-models, researching and designing a 3D-model retrieval engine the find out the required model accurately becomes a problem which need urgently dealt with. The technology of content-based 3D-model retrieval uses 3D-model's feature to set up retrieval indexes and accomplishes retrieval directly. The essence of classification and retrieval different 3D model is to extract feature from existed 3D model and describe the feature in the numerical value.In general, a complete 3D-model retrieval system contains feature extracting, feature matching, index and interface. In these aspects, feature extracting is most important, thus it becomes a key problem for resolving in 3D-model retrieval system. In the paper, the main steps and key techniques of content based 3D-models retrieval system are analyzed and summed. The spherical harmonic is employed in the paper to extract the 3D-model feature, and a 3D-model retrieval based on spherical harmonic is designed.Firstly, the research method and content of the paper are determined via summing and analyzing the international research current situation and the key technique employed by 3D-model retrieval system. The spherical harmonic is employed to extract the models feature. The advantage of the method is that spherical harmonic can help to convert the space field future vector into frequency field. This character can help to reduce the influence caused by 3D-model revolution and avoid the preprocessing of the revolution normalization. In addition, the distance based feature matching was employed to determine the similarity of two models. This matching method is fit for the feature vector extracted by the spherical harmonic based method.Secondly, we put improve the spherical harmonic based 3D-model retrieval algorithm. We sample the surface of the 3D models by the latitude and longitude method. The points on the surface of the 3D models are future processed by the spherical harmonic. The spherical harmonic is applied to calculate the feature vector from the point. We sample the 3D model's surface before extracting feature. After sampling, the spherical harmonic been used to analyze the point, and reconstruct the 3D-model's signal. Rotation invariant feature vectors are extracted. Those 3D model's feature vectors would contain more structural information of 3D-model and reduce the influence of the surface noise.Finally, the 3D models retrieval system designed in the paper is introduced. The retrieval system is a desktop system. The users retrieval the 3D models by choose a model in the database. The file management facilities of the operation system are applied to organize the 3D models file. The feature database of the 3D models is pre-calculated. In the system the standard 3D-model database has been used.