Remote Sensing Image Retrieval Based On Relevance Feedback Under Web Environment

With the rapid development of technology of remote sensing, the quantity of remote sensing image becomes increasingly large. How to get the image that users need quickly and accurately in massive database has become an urgent problem of the retrieval of remote sensing image, so it is necessary and important to develop a more accurate and efficient retrieval technology of remote sensing image. In this paper, in order to overcome the "semantic gap" problem which exist in remote sensing image retrieval currently, and adapt to the current Web-oriented software development trend, introduces a relevance feedback sensing image retrieval method based on weight adjustment and query points move, and establish a prototype remote sensing image retrieval system based on semantics under web environment, achieve intelligent remote sensing image retrieval.The main works are following in this paper:1. We study the implementation mechanisms and methods which are suitable for remote sensing image retrieval relevance feedback. First, we use the algorithm based on weight adjustment and query points move to execute relevance feedback retrieval. Second, we use interactive feedback to transfer the requirement and knowledge of users back to system to optimize the query results. This program overcomes the "semantic gap" problem which existed in the remote sensing image retrieval process to some extent.2. We study the implementation mechanism and strategy of remote sensing image retrieval network environment. Through using relevance feedback, optimizing image database, optimizing retrieval process, as well as rich application development technology based on HTML5and JavaScript, finally, we achieved the remote sensing image retrieval based on web environment.3. We design and develop a prototype remote sensing image retrieval system orienting to web environment. The retrieval method in this paper using a two-step query mechanism, and the implementation process which is from coarseness to fine, and continuous optimization. Finally, this paper analyzes the evaluation methods of different types of image retrieval algorithms, and verifies the feasibility of technical solutions and prototype systems.Experimental results show that the designed scheme has better retrieval performance, and has good application potentials.