Under the current rapid urbanization process,it is an urgent task to map the regional and global built-up urban areas timely and accurately.The DMSP/OLS night light data is one accurate,affordable and convenient dataset to reflect the urban distribution and built-up area boundaries.In order to promote the accurate level,this paper develops a Neighborhood Statistics Analysis method (NSA) for mapping the built-up area,which is based on the relative differences between neighborhood pixels.The proposed method indentifies the mutation region by raster calculating and accurately extracts the built-up area boundaries by combining the threshold method.Then,the proposed NSA method is applied to extract the built-up urban areas of China's 34 main cities in 2009 and the accuracy is validated by using the results extracted from TM data and the global-fixed and local-optimized threshold methods based on DMSP/OLS night light data.Results show that the pixel numbers of NSA-extracted built-up lands match well with those of TM-extracted built-up lands.The coefficient of determination R2 is 0.966,with root mean squared error RMSE=191.64 and relative accuracy RA=82.79％.The Landscape shape index,aggregation index,edge area ratio index and connectivity index also show highly consistent (R2=0.475 4,0.366 2,0.858 9 and 0.915 3,respectively).In addition,the proposed NSA method,which significantly overcomes the disadvantages that are associated with the global-fixed and local-optimized threshold methods,accurately maps both the large patches of built-up areas in urban regions and the small patches of built-up areas in surrounding towns.
国家科技期刊平台
NSTL
维普
万方数据
