An efficient assembly retrieval method based on Hausdorff distance

Abstract

Abstract An assembly provides significant knowledge in design reuse, since it can represent parts information and their relationships comprehensively. The retrieval of assembly has the practical value in accelerating design and enhancing efficiency. Our previous work addressed this issue by using a method based on earth mover's distance. However, it requires a longer, less favorable retrieval time to obtain the retrieval results. To solve this problem, we propose an efficient assembly retrieval method based on a modified Hausdorff distance (MHD). By employing shape distributions, each part is quantitatively described as a point and an assembly is turned into a set of points, then the MHD is performed to evaluate the dissimilarity between the generated point sets. The MHD could be used in overall retrieval of assembly, since it considers the contributions of similar and dissimilar parts. For the scenario of flexible retrieval, a directed Hausdorff distance is proposed, which only compares the local dissimilarity. Experiments are carried out to demonstrate the accuracy and efficiency of the MHD, compared with the earth mover's distance and the vector space model. Our study reveals that the proposed method can retrieve relevant assemblies efficiently on the premise of assuring the accuracy.