Study On The Extraction Of Impervious Surfaces By Spectral Index Method

With the accelerate of urbanization process,the ecological system is changing.The farmland,forest,bare soil is gradually replaced by artificial surface,which led to a series of environmental problems,such as floods,drought,environmental pollution,heat effect and so on.The increase of urban impervious surface is the most obvious character of urbanization.Therefore,extracting impervious surfaces in large areas rapidly and accurately has a great significance to study urban expansion,population growth and migration,and global environmental change.In the existing methods of extracting impervious surface,there are many problems,such as the difficulty in endmembers selection,complicated implementation process,and subjective data processing by manual,which need to consume a lot of time and energy.Compared to the above methods,index methods have several merits including no parameters,the rapid calculation speed and large scale mapping,so this paper will use the different spectral index methods to extract the impervious surface and explore the potential applications on regional scale.The specific research mainly includes the following aspects:1)A modified index(MNDISI)is proposed.The index is proposed to sharpen the thermal infrared bands to 30 meters and are applied it to NDISI.The results show that MNDISI is beneficial to the identification of urban information.2)A method for automatically setting thresholds based on generalized Gaussian model is proposed.In the process of extracting impervious surface,the thresholds usually are set by the visual interpretation and artificial selection.In order to avoid the subjectivity and time-consuming problem,we put forward the method of setting the thresholds automatically,this method has a high accuracy by the validation of comparison analysis of thresholds;3)The different indices are compared each other.In order to verify the impact from the seasonal alternate on extracting impervious surface,we use the images of the four different seasons to calculate SDI index,which shows the summer image is beneficial to the extraction of water information.Therefore,the summer images are used in the following analysis of different spectral index,the validation results show MNDISI index has a high precision in the impervious surface extraction.Meanwhile,the effect of BCI and NDISI indices are sound.4)In order to verify the effect of extraction impervious surface in large areas by index method,the Google Earth Engine(GEE)platform is used for combine the three indices of NDISI,NDVI and MNDWI to extract impervious surface in the Shangdong region.The overall accuracy of the extraction is 84.03%,the Kappa coefficient is 0.68.Through the verification and comparison of the above methods,it is found that spectral indices have a good effect on extracting impervious surface.At the same time,it is feasible and applicable to the extraction impervious surface in large areas by combining with the three indices.