Research And Application On 3D Model Retrieval Technology Based On Symmetry Feature

In recent years, improved modeling tools and scanning devices are making the acquisition of 3D models easier and lower expensive. 3D graphics hardware has become fast and cheap enough. It leads to high demand for 3D models from a wide range of sources. The 3D model data have become another new kind of multimedia data type after the image, the sound and the video. Meanwhile, the development of internet technology enables people to gain 3D models easier and more convenient, and creates the condition for spreading and sharing the resources of 3D models. As a new type multimedia data, 3D model retrieval on the base of features is proposed in user's demand. An important issue is how to help people find their desirable 3D models accurately and efficiently from the model database or from the web.The key problem of 3D model retrieval is how to effectively extract the feature of 3D model. In this paper, we study the domestic and foreign research on methods of 3D models feature extraction. We focus on the method which is based on symmetry features. Theoretical analysis and experimental results have proved that the proposed methods are feasible and effective. The main innovations of this thesis are summarized as follows:1,An enhanced reflective symmetry descriptor is proposed in this thesis. We introduce calculating methods of the reflective symmetry descriptor into projected images of 3D model, that cause the enhanced reflective symmetry descriptor not only suitable for the 3D model retrieve, but also suitable for the 2D image retrieval. This is an innovation.2,This paper presents a rotational symmetry descriptor. Under the driving of the descriptor, we combined statistical feature to search step-by-step. This method increases the retrieval efficiency which is low through single retrieval by statistical feature. This is the second innovation.The main contents of this thesis are shown as below:(1)This paper introduces the research background of the 3D model retrieval. At the same time, we review existing 3D models feature extraction technology based on the shape. And we analyze the frame of 3D model retrieval system. We also present the pretreatment technologies, the similarity matching methods and the performance indicators. (2)Shape description is the foundation of shape matching, directly affecting the accuracy of matching results. This chapter presents a description method of 3D model based on reflective symmetry feature. This method mainly obtains its view features from different viewpoint projection. Then we calculate the reflective descriptor and Fourier descriptor of images. Finally we gain the enhanced reflective descriptor by combining these two descriptors. Therefore we match arbitrary 3D model which is described with enhanced reflective descriptor by L2 distances. The method is not only suitable for 3D model retrieval, but also for the 2D image retrieval, especially for the interface of 2D sketch in 3D model retrieval system.(3) This chapter presents a description method of 3D model based on reflective symmetry feature. The statistical feature is only suitable for rough classification. So we use the rotational symmetry feature and the statistical feature to retrieve step by step. Firstly, we calculate the k-fold rotational symmetry and statistical features. Then we search M the most similar 3D models according to the rotational symmetry features. Finally we search N the most similar 3D models from M according to statistical features as the final results, M>>N.