Research On Retrieval Of Vector Graphics Based On Multi-stage Architecture And Relevance Feedback

At present, it's become more popular to use computer to draw vector graphics in the area of architecture design, mechanism manufacture and industry design. With storage technology developed, it's has many graphics databases in everywhere. How to retrieve useful information from large database becomes new research aspect in the world information retrieve domain. This paper mainly focuses on the studies of several important technologies of vector graphics retrieve: feature extraction, similarity measurement, relevance feedback.This paper presents a new method for retrieving engineering drawings. In this method, features are firstly extracted out from drawings based on geometric primitive and then a multi-stage architecture is used to retrieve the drawings. The statistical and structural features of the geometric primitive, which are used to characterize the drawings, are extracted out after the pretreating of drawings. Then the multi-stage architecture is constructed as follows: in the first stage, those simple statistical features such as the number of the primitive are used to retrieve the drawings in the database so as to reduce the retrieval range; in the second stage, the candidate drawings set got from the first stage are retrieved accurately again based on the features such as the slope of line, the radian of arc. Compared with classical retrieval methods, this new method uses both statistical and structural features so as to avoid the problem of normalization which arises from the unitary-vector characterization of two different types of features.On the other hand, it's hard to solve the problem of human recognition domain by simple or unitary feature vector. It's a useful method to query graphics by itself characteristic based on using relevance feedback technology. This paper reference some classic feedback algorithm, and it's present a new method that it could take some new technologies to traditional feedback algorithm such like features induce and weight gene.We research on the combine of two retrievals that based on architecture andrelevance feedback. This paper present the algorism that it's sufficient make use of advantage of two technologies, process some features with scientific technology, reduce features dimension as more as possible. And it has improved upgrade of the retrieval precision and efficiency of vector graphics.