The Research Of Glacial Lake Extraction Based On Landsat-8 Oli Imagery In High Mountain Region Of Asian

Glacial lake is formed by the combined effects of glaciation and warming climate during the situation where the warming global climate and the shrinking glaciers are normally in high mountain regions since the last glacial period.Glacial lake is an important water resource in plateau area.It is widely distributed in High Mountain Asia,including Altai mountains,Tienshan,Kunlun mountains,Himalayas,southeastern Tibet and so on.With the glacial lake state changes,a glacier-related disaster-glacial lake outburst flood(GLOF)will be happen when the moraine dam destroyed by the pressure of glacial lake.In recent years,the glacial lake is expanding and increasing fast with the global climate warming and snow melting.At the same time,occurrence frequency and destroyed level of glacial lake outburst increased gradually,which brings a great threat to the downstream region of the residents'life and property security.Therefore,effectively extracting the glacial lake information from remote sensing images,real-time monitoring of glacial lake,and estimate the high potential outburst risk of glacial lake,is one of the important basis for exploring the relationship between climate change and water resources and for the glacier-related disaster response to carry out the work of disaster prevention and mitigation.With the increased availability in spatial and temporal resolutions from the 1960s,remote sensing technology has shown increased probability for the identification,extraction,and monitoring vulnerable glacial lakes.Due to the large difference in the spectral characteristics of the glacial lake and the small area of the glacial lake in the remote sensing images,there is still a lack of effective glacial lake extraction method in a large scale.Therefore,based on the Landsat-8 OLI remote sensing image,this research focus on the exploration of glacial lake extraction methods,including the feature selected for glacial lake extraction,the optimized traditional threshold segmentation algorithm and the fuzzy clustering algorithm,and then this article proposes a improved C-V model(Chan-Vase model)based on"global to local"water extraction algorithm,named TSCV(threshold and simplified C-V model).In this study,the extraction algorithm of glacial lake information in remote sensing images is studied,and the following results and conclusions are obtained.1.In order to design a effective glacier lake extraction algorithm,this study selects 23 characteristics of glacial lake,including the band reflection,water index NDWI(normalized difference water index)and transformation characteristics.Experiments on feature selection were conducted by random forests and decision trees.For combining with Landsat series data and removing the correlated features,NDWI1is finally used as the water index characteristic in glacial lake extraction.2.Two classes of pixel based methods--traditional threshold segmentation and fuzzy clustering are improved and optimized,and the idea of"global to local"is introduced to extract glacial lake.The traditional single threshold segmentation method is difficult to obtain the single effective global threshold.In this experiment,the optimal threshold iteration method is introduced into the glacial lake extraction,and the optimal threshold of each water unit is determined through multiple iterations,so as to improve the accuracy of image segmentation.The fuzzy clustering method is difficult to classify each glacial lake unit accurately.Therefore,the idea of semi-supervised model is introduced into fuzzy clustering algorithm,that is,using a small number of labeled samples to assist fuzzy clustering process for improving the accuracy of classification.3.In this study,a region-based model,C-V model is introduced in the symbolic function of pressure,and retains the gradient.To simplify the model and reduce the amount of computation,this article combined the"global to local"water extraction algorithm and improved C-V model,then put forward a system of glacial lake extraction algorithm-TSCV.TSCV is a region-based model,which can effectively identify the weak boundary information according to the statistical characteristics of the region,so as to has a better extraction effect for the small glacial lakes(area less than 0.1km~2)which is easily confused with the background.In addition,TSCV has the good anti-noise ability,and some inhomogeneous regions can not get a stable contour.Some objects which has the similar spectrum with glacial lakes,such as mountain shadow,glacier,noise and so on,are often unable to obtain a closed curve.Thus TSCV can avoid the interference from these objects and further raise the glacial lake extraction precision by this characteristic.