| In recent years,with the rapid development of computer vision,3D graphics hardware and software technology,the 3D models have been widely used in CAD product design,virtual reality,3D printing and 3D products presentation,ect.On the Internet,there have been a huge number of 3D models,which is increasing as exponential type.Therefore,how to quickly and accurately search 3D model what users need has become an urgent problem to be solved,which has an important theoretical value and practical significance.To reduce the redundant views of existing research,this paper proposes a multi views clustering method based on k-means algorithm.Firstly,vertices of the C60 structure are used as the viewpoint to obtain rendering orthographic views of 3D model,Secondly,the geometric invariant moments of each view are extracted to establish the training sample.Finally,it proceed cluster analysis by selected algorithms to obtain some views on the characteristics of 3D model.This method can improve the description ability of selected views and greatly reduce the complexity of the algorithm.By using the image instead of the hand drawn sketch as input to retrieval 3D models which is more objective and accurate.Firstly,the design of preprocessing algorithms for complex background images is presented to get effective image data.Secondly,the image feature extracted by Zernike moments,DFT and CSS.Finally,to further improve the performance,the D-S evidence theory is used to fuse the single feature vector.Since human beings has viewpoint independence to real things,we chose the longest contour corresponding view as optimal view to extract feature,which is close to the user selected results.Synthesizing above all outcomes,the 3D model retrieval system is designed and implemented based on images,which provide a platform for the follow-up study. |
