Research On Projection-based 3D Model Retrieval

With the development of computer graphics and related software and hardware technology,model scanning and 3D modeling technology becoming more and more mature and a lot of 3D models were constructed.The 3D model in product design,simulation,scientific research,animation,film and television,network game,and other fields has been widely used.With the rapid development of Intertnet,the information was spread more quickly and convenient to process 3D models.It is so difficult to design and implementation of highly laborious simulation of 3D model.If users got these similar models from the shared resources on the Internet,then modify the design according to the demand.This way can significantly improve modeling efficiency and resource utilization.Therefore,Using 3D model retrieval technology to sharing and reuse have important research value.The feature extraction is key in 3D model retrieval technology.The 3D model based on projection features has a good ability to distinguish and were robutst to geometrical noise,grid degradation and subdivision.These advantages make it be concerned widely.But the local information of the model surface was obtained by the projection of the method and is difficult to extract the feature of rotation invariance.The feature is also insufficient to describe internal space distribution information of the model.According to the above problem,this paper researches on the 3D model retrieval algorithm based on projection.Three-dimensional model was obtained from different Angle of projection images using mature image features to retrieve.The main content of this paper is as follows:(1)In this paper,the 3D model retrieval methods were researched and introduce the research status.Then summaring the main steps of 3D model retrieval and 3D model retrieval system were used widely.(2)In this paper,The methods about three dimensional mode feature extraction based on image were summarized.These methods in general can be categorized according to its different implementation principle and expounds all kinds of typical algorithms and implementation process,analysis their respective advantages and disadvantages.(3)The 3D model retrieval method based on projection is further studied in our method.The 3D model retrieval algorithm was implementated by extracting feature of different 3D model direction projection in image sequence and using multiple view clock matching algorithm.It solved the problem of rotation invariant in model retrieval effectively and implements one-to-one in the projection image direction.(4)The 3D model retrieval based on feature fusion algorithm was proposed in this paper.Through calculate the distribution of information of model grid in different spherical shell and extract interior space characteristics of 3D model.Combining the projection features and feature of spherical grid distribution.We have evaluluated the result of our approach on PSB dataset,experimrnt results show that the fusion feature can describe the model better.