Design And Implementation Of Distributed Image Retrieval System Based On Improved SIFT Algorithm

With the development of the Internet, higher demands for search engines have been increased, especially the demand for image retrieval, while the existing technique performs poorly on both accuracy and efficiency, In this context, the image research group of ICT has decided to develop a distributed image retrieval system based on the improved SIFT (Scale-invariant feature transform), using it as a research platform to further enhance the accuracy and efficiency of image retrieval.This paper mainly introduces two parts, namely the improvement of SIFT and analysis, design and realization of the distributed image retrieval system based on the former.The article expounds on the research of SIFT, raising the deficiency of SIFT and the improved SIFT, i.e. by bringing in COP (Combined-Orientation-Position) optimization. We improve the accuracy of image contrast by conducting direction verification of spatial vector. Meanwhile we verify the enhanced accuracy of image recognition using the improved SIFT by comparing with other image retrieving algorithms.The paper also expounds on the demand analysis, preliminary design and detailed design of the distributed image retrieval system. Demand analysis elaborates on the system’s functional and non-functional demands. Preliminary design states the system’s structure and design. In terms of detailed design, it describes the system’s development configuration as well as the optimized algorithm of strategy module, communication module, data storage module and the complete process. We introduce testing on the system’s overall performance, experimentally justifying the distributed system can enhance the efficiency of image retrieval.Finally, we carry on simple demonstration of the system, summarize the research and raise outlooks.