Research On 3D Model Retrieval Technology Based On Multiple Feature Fusion

With the development of various forms of information,the demand for 3D models in various fields has increased dramatically.More and more 3D model forms have been created.It is a complicated and time-consuming task to re-create the 3D model that meets the requirements.How to find the model that can best meet the demand rapidly from a large number of 3D models is the current research needs.Moreover,the emergence and development of various 3D model retrieval techniques are of far-reaching significance for improving the sharing rate and reusing of 3D models.Generally,for the problems that the single feature cannot describe the model information completely,the feature point matching accuracy is not high enough,and the feature fusion method is not universal.This paper proposes a 3D model retrieval method based on multi-feature fusion,improves the feature point matching method RANSAC algorithm,and adopts adaptive feature fusion method to form a new feature description.The 3d model retrieval technology is improved from many aspects,and a multi-feature 3d model retrieval method is proposed.(1)We propose a multi-feature 3D model retrieval method based on shape context.Extract views of the model in all directions,and we combine ORB features to describe the local information and shape context features to describe the global edge information,In the process of detecting ORB feature points,we use the improved RANSAC algorithm to match the similarity of the feature points between the models,and the feature point pairs are similarly sorted.It effectively eliminates the excessive distance between the first feature point pair and the second feature point pair,improving the accuracy of feature point matching.At the same time,the anisotropic diffusion filter is applied to the image to effectively filter the influence of noise.After the Canny operator extracts the edge features,we use the shape context feature to further calculate the edge information..Finally,feature weighted summation is used to form a new feature description method.The comparison experimental results show that the method can effectively improve the retrieval accuracy.(2)We propose a multi-feature adaptive fusion 3D model retrieval method.Extract views of the model in each direction firstly.Then on the basis of extracting ORB features and shape context features,we extract the shape histogram based on the energy model.Analyze the frequency band of energy concentration and statistically obtain the shape histogram.On the one hand,the problem of inaccurate local feature points detected by the ORB feature is settled,on the other hand,it is merged with the shape context descriptor,so that the shape features of the model are more fully described.Combine ORB features,energy shape histograms,and shape context features together adaptively.We select the benchmark features,calculate the coverage of each feature,and derive the fusion weight.The experiments were compared on the two model libraries of SHREC2012 GTB and PSB,which further improved the accuracy of 3D model retrieval.