One Method of Urban Land Covers Information Extraction

Abstract

To study urban land cover information extracting method used one scene LANDSAT-TM remote sensing image of Chongqing city, China. Since NDVI calculation can enhance vegetation information, a calculation expression which is built according to water spectrum feather can enhance water information, and NDBI (Normalization Difference Building Index) calculation can enhance impervious surface information, the three calculations were used to get three thematic images, and after stacking them on to be one image, supervised classification used k-Nearest Neighbor algorithm based on Object-Oriented features of the image was used to get vegetation , water and impervious surface information. To value the final accuracy of the classification, 100 random sampling points were chosen and the high spatial resolution remote sensing image of Google Earth was taken to be the reference, the overall classification accuracy is 83.4%, and the Kappa statistic is 0.814.