A region-based method of vegetation coverage extracting in complex terrain areas using polarmetric SAR data

Vegetation coverage is an important indicator for forecasting geological disasters in mountainous areas such as landslide. However, it is a challenge to extract vegetation coverage in complex terrain from SAR image. A major problem is that the variation of the backscatter coefficient of the same object varies with the local incidence angle. As a result, a large number of discrete points appear in the classification results. Focus on this problem, a region-based method of vegetation coverage monitoring has been presented in this paper. Based on the results of yamaguchi decomposition and SVM algorithm, the Watershed algorithm is used to over-segment the image and regions are merged on a pixel-voting basis. When the modified method was applied on RADARSAT-2 data, the research results show that the completeness and correctness are improved compared with the method based on yamaguchi decomposition and svm algorithm.