A CAD model retrieval framework based on correlation network and relevance ranking

The computer-aided design (CAD) models contain abundant domain knowledge, either structure, material, or process information. An efficient retrieval ability for these reusable design resources will provide designers invaluable support for efficient product development. With this goal, this paper proposed a novel CAD model retrieval framework based on correlation network and relevance ranking. First, a multi-layer network was constructed to express the high-level local correlation between CAD models. Then, the global shape comparison method is employed to determine the CAD models most similar to the query, called the relevant subset. Finally, the relevance ranking based on the Bayesian theory can be performed by analyzing the correlation between the relevant subset and other CAD models. The relevance probability determines which CAD model is the most relevant to the query, and the ranking list can be finally obtained. Experimental results and comparisons with state-of-the-art methods demonstrate the superiority and user-friendliness of the proposed method.