AMSR-E image classification based on SVM for flood and waterlogging monitoring

This paper describes the application of Support Vector Machine (SVM) for AMSR-E image classification in order to monitor flood and waterlogging. The SVM technology can play a unique role in the AMSR-E image classification, because of the difficulty to acquire much pure type pixels as training samples in the coarse-resolution AMSR-E image. The experiment result indicates that classification image based on SVM can clearly reveal the large scale flood and waterlogging patterns over Huaihe River Basin on July 6, 2003. The classification overall accuracy is 97% more than the Neural Network method. Furthermore, SVM shows the better time-saving ability.