3D Model Retrieval System Based On Graph Matching

With the rapid development of science and technology,the 3D model is widely used in many fields such as medical,film,games and today's VR technology as a polygonal object that can visually simulate the real world.The rapid development of these technologies has led to the geometric growth of the number of 3D models.How to quickly obtain the 3D models required by users in a large number of 3D model libraries has become an inevitable trend of development.The 3D model retrieval technology emerges in this environment.The early 3D model retrieval technology was mainly text-based,but the text-based 3D model retrieval technology not only required a large number of human resources,but also the accuracy of retrieval was not high.The content-based 3D model retrieval technology is more concerned with the visual characteristics of 3D models.Compared with the text-based 3D model retrieval technology,the accuracy of retrieval is greatly improved.This paper focuses on the group graph matching method for realizing the similarity measure of the 3D model and the construction of massive 3D model retrieval system:1)In the group graph matching method,The similarity between the cluster and the cluster is obtained by calculating the local unary and pairwise correspondences in the cluster.Based on the similarity,the group graph matching is modeled as an integer quadratic programming problem.By maximizing the objective function,the correspondence between the clusters is obtained,and then the similarity between the group graphs is obtained.2)Based on the asynchronous web crawler framework Scrapy to achieve 3D model data crawling,based on the distributed system architecture Hadoop build Nosql database Hbase to realize the storage of the 3D model description information,based on the distributed search engine Elasticsearch to realize the construction of the 3D model index.3)Applying the group graph matching method to the 3D model retrieval system,constructing the graph structure of the 3D model by using the 2D view,and using the group graph matching method to achieve the similarity matching between the 3D models.4)The validity of the group graph matching method is verified by experiments on the composite graph,real graph and3 D data set.The feasibility of the group graph matching method and the robustness of the system are verified by the performance test and pressure test of the 3D model retrieval system based on group graph matching.