The whole scheme design on 3D model retrieval

Abstract

The essence of 3D model retrieval is to classify and to sort. An algorithm based on neural network is proposed in this paper. After pretreatment, six projection viewpoints are selected to project the 3D model in order to generate six two-dimensional images. Then the images are transformed by Fourier Transform to obtain the vector features. Finally, the dimension of vector is compressed to input into neural network to retrieve. The advantage of neural network is to provide an interface between the system and the users. Users can retrieve more than one times if not satisfied. Proved by experiments, the algorithm can get better performance and satisfactory efficiency.