Research On CBIR Based Product Design Patents Retrieval System

Product design level not only reflects a nation's level of industry and technology but also decides the integer quality of product directly. It's also the key factor of product's market competition. Product design patent is a kind of new patent in our country, since we just started research on this domain from 1980s. But now product design patent has gained more and more recognition from company and nation.Product design patents retrieval technology has great significant for patents protection. Most of the patents retrieval system in use today are keyword based, which does not make full use of the visual characteristic of the product design patents and cannot satisfy the requirement of the retrieval and protection for product design patents. A novel hybrid search method of product design patents is proposed. It is a feedback based method which make use of the product hex-vision images and commented keyword. The result of the experiment indicated that the proposed hybrid method can inquire and retrieval the design patents database precisely and effectively and contribute to much for the use and protection of the product design. Specifically, we make the following contributions:(1) Introduce CBIR technology to product design patents retrieval, establish a new method of product design patents retrieval.(2) Take high level semantic information and low level picture visual characters as design patent characters, the two kinds of information cut down each other's shortage.(3) Introduce a semi-auto method of semantic extraction from design patents, use relevance feedback technology to improve the semantic net.(4) A product design patents retrieval prototype system is developed. In this prototype, we implemented some of the retrieval algorithms mentioned in the thesis.In the first part of the thesis the general contents of product design and product design patent and its importance is introduced. Current design patents retrieval systems and the shortage is discussed also. The conception and study status of CBIR and its relative technology is described in chapter 2. In chapter 3 we discuss how to introduce CBIR to product design patents retrieval, what problems will show themselves and how can we resolve these problems. Chapter 4 is mainly about the implementation of PESARCH, the product design patents retrieval prototype system. In the last part the whole thesis is conclusion, and some of the relevant directions for future work is discussed.