Extraction Of Impervious Surface Based On Sentinel-2 Data And Its Realization By IDL Programming

The urban spatial expansion have replaced the vegetation-dominated natural landscape with various impervious surfaces.This replacement has brought about negative impacts on the global ecological environment and has raised public awareness of evaluating ecological environment indicator.So it is very significant to gather the accurate impervious surfaces information quickly.Today’s remote sensing technology has provided numerous algorithms for impervious surfaces by scholars from various countries.Among them,the index method is widely used because of its simplicity,operability,and high accuracy.However,there have less research that applied it to the Sentinel-2 data currently,especially for the in-depth comparison of different impervious surface indexes.Therefore,this paper taking Fuzhou,the capital city of Fujian Province,Nanchang,the capital city of Jiangxi Province,Nanjing,the capital city of Jiangsu Province China,as the study areas,four pairs of two date-coincident images of Sentinel-2 and Landsat 8（two pairs of Fuzhou acquired on 16 September,2017 and 6September,2019;one pair of Nanchang acquired on 1 November,2017;one pair of Nanjing acquired on 6 June,2018）were selected as the data source.Based on the literature research,eight impervious surface indexes（including:NDBI,ISA,IBI,BCI,EBBI,CBCI,ENDISI,and NDISI）with more skilled and widely used are selected for the extraction of impervious surfaces and compared the performance of impervious surface extraction among them.On this basis,MHUTS method and NDVI index were added into the NDISI index to propose the MNDISI_S2-L8 index which based on two date-coincident images of Sentinel-2 and Landsat 8.The study uses two date-coincident images of Sentinel-2 and Landsat 8 dated on 16 September,2017 in Fuzhou,as an experimental data,to compare the performance of impervious surface extraction among the above nine indexes from the aspect of qualitative and quantitative analysis.Finding that MNDISI_S2-L8、ENDISI、NDISI and IBI were more suitable for Sentinel-2 data,especially MNDISI_S2-L8 performs best.In addition,the comparison based on the retrieved impervious surface information among the four indexes was performed.The different images of Sentinel-2 were used to investigate whether the four indexes could obtain high accuracy in retrieving impervious surface.The three test regions were selected from Fuzhou（6 September,2019）,Nanchang and Nanjing of China,with date-coincident images pairs of Sentinel-2 and Landsat 8.The comparison showed that the four indexes were all suitable for the three study areas,and MNDISI_S2-L8 has the best performance than other indexes,the secondly was ENDISI.The result showed that MNDISI_S2-L8 was used to retrieve impervious surface information when the two date-coincident images of Sentinel-2 and Landsat 8 were acquired.With Sentinel-2 data obtained only,the ENDISI index can be used for extraction of impervious surface information.So that the higher spatial resolution（10m）and higher accuracy impervious surface images were obtained.At the same time,according to the IDL development language and the secondary development technology of ENVI,a rapid-extracted impervious surface system based on Sentienl-2 data was developed,this system has an easy to operation and friendly user-interface,which is convenient for the following same research.