To reuse 3D CAD models more effciently, a new 3D CAD model retrieval algorithm based on Bag-of-Features is proposed. Firstly, a large number of local feature points are extracted from the surface of 3D CAD model, and every feature point is associated to a local shape descriptor using a matrix. Then, the 3D CAD model can be represented as a set of local shape descriptors. Secondly, the local shape descriptor is mapped to a visual word from the visual dictionary according to the minimum distance. And the 3D CAD model is described by a histogram of occurrences of these visual words. Lastly, the L1 distance metric method is taken to compute the similarity between the two histograms of occurrences of visual words, which can give the similarity coeffcient for two compared 3D CAD models. Experiments results show that the algorithm can effectively support 3D CAD model retrieval, and the effciency meets the requirement of engineering application. © 2014 Binary Information Press.
