The Use Of Remote Sensing Information Extraction Research, Land-based Multi-source Data

Land use information extraction from remote sensing image is one of the hot area of remote sensing applications. With the upcoming high- spatial resolution images, the research and applications on the land use information remote sensing extraction is highly focused on.In this paper, We analyzed the landscape area from LongQiao town, ChengDu city with the data source from BJ-1 image, SPOT-5 image, QuickBird image and landscape map of LongQiao town. First of all, we preprocessed the source data images with a series of methods, such as geometric correction and registration , multi-source remote sensing data fusion,(fusion of BJ-1 panchromatic image and multi-spectral images, fusion of BJ-1 image and SPOT-5 image), etc, then, we visually interpretated land use types of LongQiao town by QuickBird image and landscape map of LongQiao town, and, with the government newly issued "Present Situation of Land Use" and ground feature of LongQiao town, established land use classification systems of LongQiao town, then, we extracted the land use information from the multi-source remote sensing images with supervised classification method, at last, we extracted information from the shape knowledge-based method from the classified road and village, analysed it and made the following conclusions:(1) Geometrically corrected the landscape map of LongQiao town, BJ-1 panchromatic image , BJ-1 multi-spectral images and SPOT-5 image with normal polynomials, and acquired satisfactory results due to the flatness of the landscape of LongQiao town.(2) After BJ-1 panchromatic band image and multispectral band image using HIS method, both the panchromatic high spatial resolution and the magnitude of multispectral band information have been acquired.After the fusion with BJ-1 panchromatic band image and green/red/infrared band of SPOT-5 image, the result shows to be more colorful, and the details of BJ-1 panchromatic band image is preserved. while compared to the SPOT-5 image, noise level is higher than the original SPOT-5 image. (3) Using the same samples and the same classification method extract the land use information with multi-source image datas at the same area, the result differs significantly: the classification accuracy of BJ-1 image is the lowest, the classification accuracy of BJ+SPOT image is higher, and the classification accuracy of SPOT-5 image is the highest.(4) The extraction accuracy of road and village information after the shape knowledge-based method is greatly improved.