Local Scale-Based 3D Model Retrieval for Design Reuse

Xiao-Jun Zheng,Yi-Shou Wang,Hong-Fei Teng,Fu-Zheng Qu

DOI: https://doi.org/10.1007/s00170-008-1701-4

2008-01-01

Abstract:3D model retrieval plays an important role in product design reuse. We present a local scale-based model retrieval methodology for 3D mechanical parts library. On the grounds that most of the mechanical parts have visible main planes, we firstly propose a novel approach, i.e., the plane comparison-based pose normalization. This approach was employed complementally with principal component analysis (PCA for short) on the normals of the model's face to perform rotation invariance. After pose normalization, the model is scaled locally along a certain axis so that the size of its local features can be adjusted reasonably. In this way, the feature vectors of the scaled model can represent all the local features. The feature vectors of the deformed model are extracted by the real spherical harmonics. Finally, the feature vectors and the scale factors are used in an experiment for similarity comparison. Experimental results showed that the proposed method outperformed five other methods on the National Design Repository dataset and our own complex mechanical parts dataset.