Permanent snow and ice extraction based on object-oriented classification in Himalaya

As the highest mountain chain in the world, the Himalaya´s snow and ice are melting gradually. How to extract the border and area of permanent snow and ice in Himalaya becomes a very important problem for global climate change research. In order to explore methods to extract permanent snow and ice in unreachable area, this paper takes the multi-source remote sensing data as the basic of multi-scale segmentation, and utilizes the decision tree to probe object-oriented permanent snow and ice extraction. This paper adopts the vector distance method and optimal segmentation scale calculation model to solve the problem of segmentation scale selection. It takes horizontal and vertical distance between the boundary of image objects area after segmentation and the actual boundary of classification targets as the precision difference index, and estimates the effectiveness of segmentation results in order to determine the optimal segmentation scale. Compared with the land-use visual interpretation data of Institute of Mountain Hazards and Environment, the accuracy of extraction results can reach 92.5%. It shows that the proposed methods are feasible, and the results are also credible and accurate.