Content-based three-dimensional engineering shape search

In this paper, we discuss the design and implementation of a prototype 3D Engineering Shape Search system. The system incorporates multiple feature vectors, relevance feedback, and query by example and browsing, flexible definition of shape similarity, and efficient execution through multidimensional indexing and clustering. In order to offer more information for a user to determine similarity of 3D engineering shape, a 3D interface that allows users to manipulate shapes is proposed and implemented to present the search results. The system allows users to specify which feature vectors should be used to perform the search. The system is used to conduct extensive experimentation real data to test the effectiveness of various feature vectors for shape - the first such comparison of this type. The test results show that the descending order of the average precision of feature vectors is: principal moments, moment invariants, geometric parameters, and eigenvalues. In addition, a multi-step similarity search strategy is proposed and tested in this paper to improve the effectiveness of 3D engineering shape search. It is shown that the multi-step approach is more effective than the one-shot search approach, when a fixed number of shapes are retrieved.