Abstract: Flood disaster will cause huge losses to local society and economy, timely and rapid monitoring of flood range is of great significance in disaster relief. Synthetic aperture radar (SAR) can provide support for all-day, all-weather and large-scale flood monitoring due to its active microwave imaging mechanism. This paper takes GF-3 satellite images as data source, and extracts 138 image texture features of GF-3 based on six texture description methods, such as the gray level co-occurrence matrix (GLCM) and the local binary pattern (LBP), etc. Then, the texture features with high importance scores are selected for water information extraction by using the index importance evaluation function of the random forest (RF) algorithm. Finally, the initial water extraction results are post-processed combined with mathematical morphology to evaluate the flood disaster near Chaohu Lake in Anhui province. The experiments show that the water extraction accuracy of the proposed method is better than the results of traditional threshold method (Otsu) and classification (KNN and SVM) algorithms. And the proposed method can effectively extract the influence range of flood disaster, and provide a reference for selecting appropriate texture features of SAR images for rapid monitoring of flood range.

Key words: flood disaster, texture feature, random forest, GF-3, Chaohu lake