Coastal wetland automatic extraction based on remote sensing spatial — Spectral (Tupu) coupled information

Because of the complex, diverse and varied of spectrum of wetland and spectral confusion with other land cover classes and among different types of wetlands, it is difficult for Wetland information extraction automatically. In this paper, we proposed a new hybrid approach for coastal wetlands automatic extraction based on “spatial & spectral” or “Tupu” coupled information. Due to the coastal wetland have some distribution discipline, it is depended on water and has spatial relationship with typical natural features, like coastline, lake, river and etc. So, firstly, we extract estuaries, coastlines, lakes and other features by the spectral characteristics. Secondly, we retrieved ground moisture via the TVDI algorithm. Then, Using a global spectral clustering algorithm, convert the image pixel-level into object-level. Through the function of ground humidity index classified the potential and possible coastal wetland distribution. Finally, through the spatial dependence relationship analysis with the coastline, lake, river and etc, we revised coastal wetland classification result. By this way, we got the ultimate coastal wetlands information. Experimental results show that the method can accurately extract wetlands from RS image. The classification precision can meet the application requirements.