Bulletin of Surveying and Mapping ›› 2023, Vol. 0 ›› Issue (11): 122-127.doi: 10.13474/j.cnki.11-2246.2023.0339

Previous Articles     Next Articles

Subpixel water extraction method for Sentinel-2 image

XIONG Longhai1,2, HE Yingqing1,2, LIU Momo1,2, LI Jun1,2   

  1. 1. Key Laboratory of the Pearl River Estuary Regulation and Protection of Ministry of Water Resources, Guangzhou 51061 l, China;
    2. Pearl River Water Resources Research Institute, Pearl River Water Resources Commission, Guangzhou 510611, China
  • Received:2023-08-08 Online:2023-11-25 Published:2023-12-07

Abstract: Combining Sentinel-2 image and other high-resolution satellite data, it is of great significance to monitor water resources factors such as water surface rate, water storage and ecological flow in long series, high frequency and large range. In order to improve the accuracy of water extraction and solve the problem of spatial scale effect when extracting water from multi-source moderate-and high-resolution satellite data, a subpixel water extraction method for Sentinel-2 image(SWES) is proposed in this paper. First, RWI is used to extract pure water pixel, then the expansion algorithm is used to extract the mixed water pixel. Finally, in order to solve the problem of intra-class spectral changes of surface objects, MESMA considering spatial information is used to solve the water abundance in the mixed water pixel. The results of the three experimental areas all showed that SWES achieved a good effect, the average RMSE was 0.147, and the water extraction effect is better than automatic subpixel water mapping method(ASWM), especially in the pond aquaculture area with more mixed pixels of water and land. The water area obtained by SWES in experimental area also has high accuracy, with the average relative error of 8.03%, which is lower than the 20.23% of ASWM. The results show that SWES can effectively improve the accuracy of water area extraction.

Key words: subpixel, water extraction, mixed pixels of water and land, multiple endmember spectral mixture analysis, Sentinel-2 image

CLC Number: