3D Model Similarity Calculation Combining Hopfield And Chaotic Particle Swarm

With the extensive application of digital modeling technology in field of 3D modeling and the expansion of computer network.The number of reusable and shared models in 3D CAD model library is also increasing rapidly.A modern project is usually implemented with a collaborative effort of engineers from different fields and disciplines.Since engineers come from different fields,they use 3D CAD models from their fields to design parts in the product.If the existing 3D CAD models can be reused in process of designing a new product,designing budget of the product can be effectively reduced and the development efficiency can be improved.Therefore,how to retrieve existing models in 3D CAD model library and make them reusable has become a hot issue now.This thesis mainly researches basic principles of Hopfield neural network,ant colony optimization and particle swarm optimization,and improve them.Hopfield neural network,ant colony algorithm and particle swarm algorithm are used to calculate the similarity between two models.It is concluded that Hopfield and chaotic particle swarm optimization can measure more effectively the difference between two models.The main contents of this thesis can be divided into the following parts:Firstly,this thesis introduces some advanced methods of model retrieval at home and abroad.The methods include 3D model semantic information,adjacent-topology diagram,view and distribution matrix.After research results at home and abroad are compared,the existing problems in similarity calculation and retrieval of 3D CAD models at home are summarized.Secondly,several ways to describe features of 3D CAD models and their extraction methods are introduced.Based on attribute adjacency matrix,geometric similarity between 3D CAD models is calculated by the difference of edge number of face in model.Spatial structure similarity between 3D CAD models is calculated by spatial structure of face in model.Based on geometric similarity and spatial structure similarity,global similarity matrix of two models is constructed.The optimization algorithm is used to search global similarity matrix and obtain the sequence of optimal face matching pairs between two models.Global similarity of two models is calculated according to the sequence of optimal face matching pairs.Thirdly,principles of Hopfield neural network,ant colony optimization and particle swarm optimization are analyzed.The complementary principle of neural network and particle swarm optimization in fusion algorithm is introduced in detail,and the algorithm is optimized.Experimental data is compared.At the same time,advantages and disadvantages of the algorithm are analyzed.Experimental results show that compared with Hopfield neural network,ant colony optimization and particle swarm optimization,the proposed method in this thesis can measure the similarity and difference between two models more accurately and effectively.