Content-based Remote Sensing Image Database Retrieval And Realization

As the number of remote sensing satellites increasing, and hyper-spectrum, high-resolution, polarization and other new sensors developed, remote sensing image has a bigger number, but the remote sensing image retrieval technology goes slowly. So, how to find the needed image from tremendous remote sensing image databases correctly and fast is valuable. For the need of management and retrieval of increasing remote sensing images, this paper describes research content and development of Content-Based Image Retrieval(CBIR), analyzes the actuality and problems of remote sensing image retrieval from database, research the relational method and arithmetic for CBIR。First, this paper discussed database technology, feature extraction, measurement of comparability, correlation feedback, and research feature extraction of color, texture and shape. Give an improved color coherence vector, and a shape descriptor with little-area removing.Second, promoted a"feedforward control for measurement of comparability"method, using an obvious feature to affect the measurement of comparability, reducing user's input and making the system to be more intelligentized. The method was compared and estimated.Last, detail description of my own system.