Automatic Extraction of Lakes on the Qinghai-Tibet Plateau from Sentinel-1 SAR Images

As a link between the cryosphere, hydrosphere, biosphere and the atmosphere, lakes in the Qinghai-Tibet Plateau can indicate global climate change. However, due to the remoteness and inaccessible, the measured lake data is challenging to obtain. In this paper, a new automatic water extraction method for Qinghai-Tibet Plateau lakes from Sentinel-1 SAR images has been proposed. For solving the problem that conventional methods unable to satisfy the description of large-scale statistical characteristics, we present an easy-to-implement classification method combining random forest (RF) and multi-features including a new statistical feature called object-based generalized gamma distribution (OGΓD). According to several experiments, we demonstrate that our method is capable of accurately extracting lakes, performing an overall accuracy of 98.60% and 97.06% in Zonag Lake and Yanhu Lake respectively, without the requirement of too much preprocessing or postprocessing steps at the same time.