Abstract: It is an important task to predict and prevent debris flow and its impacted areas in advance. In this paper, RS and GIS technology were adopted to predict the potential debris flow. Taking advantage of the spatial resolution of domestic high-resolution images, using NNDiffuse and Gram-Schmidt methods to realize the fusion of remote sensing images into research data, combining with support vector machine (SVM) and dynamic clustering based on soil brightness index (ISODATA), the natural surface coverage and human impact area of debris flow formation region are identified and extracted, and then the formation region of debris flow is predicted by using the spatial and attribute relationship of hidden gully and catchment. The experiment shows that different fusion methods will affect the result of debris flow extraction. NNDiffuse fusion method has the best overall effect in this paper; SVM method has the best effect, the prior knowledge has the significance in the prediction of forming region, ISODATA method without prior knowledge has better performance in the identification and prediction of debris flow, there is potential application prospect in the future.