Water Body Extraction using GF-3 Polsar Data - A Case Study in Poyang Lake

Abstract

Poyang Lake is the largest freshwater lake in China and one of the most important wetlands in the world. Chinese Gaofen-3 (GF-3) was the first C-band polarimetric synthetic aperture radar (PolSAR) with the resolution up to 1m. With its polarimetric imaging capability, 2 scenes of GF-3 PolSAR data (two adjacent scenes) were used to extract open water in Poyang Lake. After pre-processing, the polarization characteristic for different land covers in the study area were analyzed. Then these 2 scenes of GF-3 data were mosaiced together. The H/A/a-Wishart classification method was used to classify the water body out for the GF-3 data. Meanwhile, the object-oriented (OO) segmentation method was used to segment the double/volume/surface component images from Freeman-Durden polarimetric decomposition. Then the Random Forest (RF) classification method was used to classify the image to extract open water. Comparing the results from these two classification methods, we can find the advantages of the combination of the OO segmentation and RF classification method in the spatial continuity of classification results over the R/A/a-Wishart classification method.