Feature Extraction Of3D Model Based On SIFT

With the development of computer graphics,3D model acquisition technologyand the improvement of computer hardware technology, three-dimensional model notonly have made a leap in the number of growth, and is widely used in people’s life,work and so on various aspects. But to create a three-dimensional model of highfidelity requires a lot of time and effort, sometimes only need to modify existing3Dmodel of the local, can get the implementation model of reuse, thus, how to retrieveand requests the same or similar three-dimensional model has become very important,especially in recent years has become a research hotspot.The precondition of3D model retrieval is to extract the informationcharacteristics of3D model. Information is characterized by description of theoriginal3D model, it is able to describe the geometric structure of the internal3Dmodel, and3D model can describe the shape of the topology. In the3D modelretrieval technology, usually the characteristics of the model to establish an index, andthen through the index to retrieve the model to meet the needs of users.So theextracted features of good or bad is directly related to the efficiency and accuracy of3D model retrieval, and how to quickly and efficiently extract the informationcharacteristics of3D model is one of the3D model retrieval technology key anddifficult.By Lowe, SIFT features due to its good scale invariance and rotation invariance,strong ability to resist noise, less affected by illumination change, has been widelyused in image matching, image retrieval, and other fields. However, the application ofthe SIFT algorithm is not widely applied in the field of3D model identification. Butin fact, this algorithm can effectively resist image scaling, rotation, noise andillumination changes, etc, it can satisfy the3D model to identify some of the practicalneed. Due to the3D model parameterization is a very complicated problem, at the same time due to the3D surface with arbitrary topology, makes some widely used inthe2D image method cannot direct marketing in the field of3D.This paper mainly studies the3D model based on SIFT feature extractionalgorithm is proposed. At first, this paper puts forward the view with thecharacteristics of the grid algorithm, and the main steps for the model preprocessing,view projection and SIFT feature extraction. This algorithm not only on the modelsurface texture mapping SIFT feature extraction, and joined the grid characteristics,forming a combination characteristic, the combination of characteristics of theoriginal model can be fully described. The experimental results show that theproposed algorithm can effectively solve the3D model but its texture similar to thatof different and cause can’t feature matching problem. Then proposed in this papercombined with the characteristics of the spherical parameterization algorithm, thisalgorithm is used to split spherical parameterization method of three dimensionalmodel for spherical images, and then calculate the sphere and the mappingrelationship is octahedron, and expand to graphic images, to SIFT feature extraction,this method can model completely topology to the spherical surface, compared to thedirect do view the surface of the projection can provide more abundant information.This method is a kind of3Dmodel by using spherical image feature extraction. Finally,combining with the characteristics of the spherical parameterization algorithm for themodification, is one of the spherical do view projection image directly, then SIFTfeature extraction. In this paper, the algorithm theoretical analysis and experimentalverification, and discusses the use of SIFT feature extraction algorithm of3D modelof the feasibility and advantage.