Mapping of small water bodies with integrated spatial information for time series images of optical remote sensing

Small water bodies and their temporal changes are, especially in urban areas, closely related to people’s daily life they have an impact on the living environment. For small mapping with optical remote sensing images, it is challenging establish a balance between reducing incorrect detection increasing integrity of extraction. time-series application, variability spectral information also for widely used threshold methods, which are frequently solely based analysis. In this work we propose spatial information-integrated (SWM) method achieve complete accurate extraction change monitoring complex environments. Our strategy make use contextual account indistinguishability information. The roughness index calculated enhance contrast other thematic classes eventually present imagery. proposed SWM was applied different indexes automatic determination. We tested effectiveness algorithm using Landsat Sentinel-2 multispectral data from three environments (Shanghai, Guangzhou, Wuhan China) include variety river courses lakes. Nantan Lake China selected as representative experimental area test by generating inter- intra-year results. Overall accuracy (OA), F1 score (F1), producer’s (PA), user’s (UA) were quantitatively evaluate algorithm. Compared four state-of-the-art methods (supervised random forest classification, hierarchical clustering, multi-band threshold, modified normalized difference (MNDWI)), achieves better performance. found be extracted more completely extractions alleviated. overall average approx. 97% (OA) 0.95 (F1). long-time sequence (from 2007 2021) short-time interval (monthly results year 2020) agree well ground truth data. mean absolute deviations body significantly smaller than that Global Surface Water dataset developed European Commission's Joint Research Centre (3.5% vs 49.0% annual 6.0% 35.0% monthly results).