Water monitoring using single SAR image: Semi-flood area

Hydrographic information is very important. The all-weather and penetrability capabilities of SAR make it useful for water monitoring. Previous research mainly focus on delineate the water boundary, to treat the target as water or no-water area. However, there are some places they ignored, in this paper we call it as semi-flooded area, like lakebed, areas after waters receded and semi-flooded farmland, even some special situations like ruffled water surface. Here we use the maximum between-class variance method (Otsu) combining the characteristics of water backscattering coefficient to define the targets as water, semi-flooded and no-water area in the image. The result shows this method is much better than traditional threshold classification.