Research On Key Techniques Of Content-Based Optical Remote Sensing Image Retrieval

Remote Sensing Data are basic data in digital earth project. Their quick browsing and efficient retrieval are important for the extraction and sharing of remote sensing information. As an efficient means for the management and utilization of the imformation in image database from the viewpoint of comprehension of image content, the content-based image retrieval (CBIR) technique has become one of the most active research points in image databases, computer vision etc. And the CBIR technique has been a key technique for information highway and digital library projects. CBIR also provides a new chance to solve the problem of information extraction from large remote sensing image database. However, the diversity and complexity of remote sensing images and the enormous data volume bring big challenges for the effective retrival of imformation from remote sensing image databases. Content-based retrieval of remote sensing database is a hot topic by integrating multiple disciplines including remote sensing image processing, image databases, computer vision and pattern recognition etc. and has gotten international considerations. Therefore, researches on it have important meanings in theory and practice for promoting remote sensing information acquisition and sharing.This dissertation proposes some new ideas and methods on key techniques for the content-based retrieval of optical remote sensing image database. The efficiency and practicability of these new methods are validated through theoretical analysis and simulations. The main work and innovations can be concluded in five aspects as given bellow:1. A retrieval model for remote sensing images is constructed, which is based on the MPEG-7 standard. Using this new model, a seires of operations are introduced to describe the retrieval process of remote sensing images perfectly.2. An evaluation method for the adaptability of texture features in remote semsing image retrieval is proposed. The adaptabilities for five commonly used texture features extraction methods are investigated by analyzing the index rates and precisions for various remote sensing images with different resolutions and physiognomies. The calculation complexities of these five methods are compared.3. An interactive genetic retrieval method is proposed. The chromosome coding and the genetic operations are special designed by using the characteristics of remote sensing images. The new method exhibits the combination of human computer iteraction and genetic algorithm effectively. 4. An interative fractal image compression method based on muti-band wavelet transformation is proposed. Simulations show the efficiency of the new method.5. According to the achievements above, an experimental system for the CBIR of remte sensing images is designed and implemented. The system can carry out conventional CBORSIR and can also implement human-machine interaction by interactive genetic algorithm.