3D engineering model retrieval algorithm based on multiple features and similarity calculation optimization

This paper proposes a retrieval algorithm based on the binary particle swarm optimization (PSO) and the joint distribution including the distance between every two random points and the normal included angle cosine in order to improve the retrieval accuracy of 3D engineering triangular mesh model. Firstly, numerous sample points on the surface of the model are randomly chosen. Next, the distances and the cosine values of the normal angles among the sample points are calculated. Finally, a two-dimensional grid with the distance and the cosine value as the coordinate axes is established. The joint distance-cosine shape distribution matrix of the 3D model is constructed through the statistic data of sample points acquired in each mesh, using the distance L₂ between distribution matrixes to represent similarity between models. In order to demonstrate the different influence of shape distribution elements on the similarity in 3D models efficiently, binary PSO is employed to ameliorate the similarity computing process. Experimental results showed that the approach could improve the retrieval accuracy of engineering mesh models effectively.