Research On 3D CAD Model Retrieval Algorithm

The 3D CAD model virtually displays the shape and structure of products and components,and is widely used in the field of product design and manufacturing.Especially with the development of digital twins and virtual reality technology,the significance of 3D CAD models is even more significant.Content based 3D CAD model retrieval takes a 3D CAD model as an input,and finds other models with the same or similar structure,which plays an important role in reusing design results and processing methods and grouping parts.This paper researches the content based 3D CAD model retrieval algorithm.The specific work includes:First,a 3D CAD model retrieval algorithm based on the combination of global and local similarity is proposed.This algorithm aims at the existing retrieval algorithms based on the maximum clique of attribute adjacency graphs,which only focus on the similarity of the local structure of the two models,and the problem of high computation cost and low retrieval efficiency of the maximum clique.A two stage retrieval strategy is adopted.In the first stage,an overall surface line distribution representation of a 3D CAD model based on the TF-IMF(term frequency-inverse model frequency)vector is proposed;in the second stage,an ant colony algorithm is used to solve the maximum clique.The algorithm can consider global and local similarities,and reduce the calculation range of the maximum clique algorithm.Experiments show that the algorithm can effectively improve NDCG and NN retrieval indexes and retrieval efficiency.Secondly,A 3D CAD model retrieval algorithm based on autoencoder neural network is proposed.For the above algorithm,in the first stage,only the model surface line distribution is considered,and the connection relationship between the surface lines is ignored.A autoencoder neural network is used to take the attribute adjacency matrix of the 3D CAD model as input,consider the matrix element value(surface line type)and element position(surface line relationship),and integrate the TF-IMF vector to consider the surface line distribution.Experimental results show that the algorithm has an NDCG of 91.18% and a NN of 70.90%,and can further improve retrieval efficiency.Finally,a 3D CAD model retrieval system is designed and implemented based on the above two algorithms,and the retrieval results are displayed visually.