Research On Automatic Extraction Of Debris Flows With Domestic High-resolution Remote Sensing Images

Geological hazards happen continually in China. Influenced by the special geological structure and human activities, Tibet is one of the most affected areas in China and develops all kinds of geological hazards, causing enormous economic losses, while also severely restricting the economic development in this region. With the rapid development of domestic satellite in recent years, more and more reliable domestic satellite data becomes applicable for geological hazard surveys, and is playing an increasingly important role compared with traditional field survey.Relying on the project of Research on Pragmatic Techniques of Remote Sensing on Geological Surveys in Perilous and Complex Areas in Western China, this paper studied on the application of domestic satellite data in automatically extracting debris flows. Firstly, on the basis of knowing well about the characteristics of debris flows and taking advantage of its spectral features, this thesis selected three components: NDVI, NDSI, and PC1 and identified candidate regions of potential debris flows with an automatic threshold selection approach. Then we extracted the valley range on which debris flows relied using ENVI+IDL and DEM data. At last, this thesis screened the polygons using slope, area and basin factors, and ultimately identified potential regions where debris flows are liable to occur. Gyaca county, Shannan Prefecture, Tibet was selected study area, and 78 polygons of potential debris flow regions in study area were extracted with GF-1 satellite data.To make a contrast, visual interpretation is also necessary. Based on features of GF-1, such as relatively high spatial resolution, accurate positioning and large span and so on, images of eight scenes in the survey region were preprocessed including orthorectification, image fusion, mosaic and geometric precision correction. Particularly, during the process of ortho-rectification, this paper adopted the method of satellite image block adjustment without GCPs to overcome the difficulty of having no highprecision control point data in high altitude and undulate regions, and finished this step with relatively high accuracy and finally generated ortho images which were used for visual interpretation. On this basis, interpretation keys used for debris flows were established and 80 debris flows were identified in Gyaca county.Compared the results of automatically extraction with visual interpretation, together with field surveys, it could be found that the ranges of 53 debris flows and the boundaries of most debris flows in two results are matched. The accuracy of automatic extraction reached 67.9%. By analyzing the results, it can be concluded that the automatic extraction method proposed in this paper is feasible and effective and can figure out relatively precise results.