Shape Retrieval Method of 3D Models Based on Shape Distribution Graph and BP Neural Network

XU Jing-hua,ZHANG Shu-you
DOI: https://doi.org/10.3785/j.issn.1008-973x.2009.05.018
2009-01-01
Abstract:A shape retrieval method of 3D models was put forward to realize the reuse of existing design cases and satisfy the need of hierarchy similarity searching of design primitives.On the basis of the B-Rep data structure of 3D models,the discrete points set is obtained through constructing the discrete surfaces of solid models.The elements of the points set are selected at random via seeds which belong to normal distribution.The Euclidean distance sequence is acquainted by calculating the Euclidean distance between arbitrary two points from the selected points.The shape distribution graph(SDG) is then obtained via the quick sorting based on the divide-and-conquer algorithm and the distribution possibility statistics of the Euclidean distance sequence,then the SDG is defined as the shape retrieval index after robustness analysis.The nonlinear mapping of the SDG of 3D models is achieved by means of BP neural network.The shape similarity between solid models is obtained via the comparison between the SDGs,and the retrieval of design primitives is realized.A prototype system was developed to verify the proposed method focusing on the design primitives with different granularities of mechanical parts.The test results proved that the computability and precision of the method are favorable enough for practical requirements.The method has important significance to realize the design reuse and improve the design efficiency.
What problem does this paper attempt to address?