Urban Surface Water Extraction Based on Sentinel 2 Images

Water is a fundamental element of urban ecosystems, and water mapping plays an important role in urban water management and disaster prevention and early warning. There are a large number of small water bodies in the urban environment, and there is spectral confusion between water and many complex features in the city, making identifying urban surface water bodies a significant challenge. Spectral mixture analysis (SMA) is widely used in sub-pixel level urban environmental analysis. In order to extract urban water bodies with higher accuracy, this paper first combined scatter plots transformed by minimum noise fraction (MNF) with image-based manual selection to optimize endpoint selection. Secondly, due to the significant variation of spectral curves of urban surface water with spatial location, pure water pixels in the vicinity of the target pixel were retrieved using window algorithms as reference endpoints. Finally, the SMA technique was applied to estimate the abundance of mixed pixels. The results are compared with the extraction accuracy of normalized difference water index (NDWI), improved normalized difference water index (MNDWI), and traditional mixed pixel decomposition method. The experimental results show that this method has a higher recognition accuracy for water pixels, with improvements of 10.8%, 8.7%, and 4.4% compared to the three methods, respectively.