Abstract: It is of great practical significance to lucubrate characteristics of debris flow on remote sensing image. The study converts high-resolution remote sensing images of different watershed regions of debris flow from spatial domain to frequency domain. Based on the spectrum analysis and frequency domain filtering, the study establishes correlation between the spectral features and partition features of the debris flow globally and locally, and identifies different watershed regions in the high-resolution remote sensing image of debris flow, finally texts it with support vector machine. The results show that making full use of frequency domain features can effectively analyze and identify different watershed regions of debris flow in addition to the spectral characteristics, and the identification accuracy is up to 95%. Thus, we can qualitative and quantitative analysis the formation process of debris flow and different areas, providing scientific basis for the prevention and control of debris flow.

Key words: high-resolution remote sensing image, debris flow, spectrum analysis, frequency domain filter