Research On Fast Geometry Processing For Magnitude Optical Satellite Remote Sensing Image

Due to the appearance of high resolution satellite remote sensing images, it is possible now to observe the details of surface changes in a larger scale, make large-scale mapping and monitor impact of human activities on the environment. Until now it has been applied in a range of fields, such as environment monitoring, urban planning, topographical, cadastral survey, and agriculture. Remote sensing image rectification, register and mosaic are essential in all the application of remote sensing. However, the diversity and complexity of high resolution satellite images make the traditional geometry processing technology no longer applicable, which includes:(1) a wide range of imaging model makes high resolution geometric correction more difficult; (2) data volume being growth dramatically results in dealing with problems of remote sensing data in large volume; (3) design and implementation of automate geometry processing system. Based on the above issues, this article try to solve problems faced in automatically geometry processing of high resolution satellite image, which includes system design, system implementation and experimental design with verification.Automatic geometry processing of high resolution image contains ortho-rectification, automated registration and automatic mosaic. This article first study ortho-rectification methods of different sensor model, and do ortho-rectification by choosing suitable model according to satellite imaging system; There will still be a little bit of error in the geometric coordinates of ortho-rectified images. So this article uses template matching method match exactly the two images. In order to improve the control point for the accuracy, a two-step-filtering method is adopted to select the best control points. After color adjusting of the two registered images, we mosaic them into one whole image.For massive data processing, this article uses a unified block tiered storage architecture. Memory-mapping and tiled-image processing are adopted together to improve the efficiency. And then an automatically geometry processing system of massive high-resolution remote sensing image is designed, and is implemented using c++language in Visual Studio 2005 development environment.After system implementation, we tested different models depending on the imaging model of satellite data on the system, for example, IKONOS remote sensing image is used for universal imaging model; the SPOT high resolution satellite images is used for physical imaging model. For automatic registration and mosaic testing, multiple CBERS-02 2-level data products and Landsat 7 ETM+L1G data products is used. The results show that this algorithm process can implement remote sensing data geometry processing automatically and satisfy accurate registration requirements of a variety of remote sensing images, meanwhile it achieve a breakthrough in the amount of data volume in the process.