Research On Glacier Extraction Methods Based On Multi-Source Remote Sensing Data

Extracting glacier information is the global hot spots of research glaciers melting,climate change and other aspects. Remote sensing technology becomes necessary glacier extraction techniques. The rapid development of remote sensing gradually formed a multi-source characteristics of remote sensing data. Therefore, the study of based multi-source remote sensing data glaciers extraction method can exist multi-source remote sensing data utilization to improve the accuracy of the extracted glacier.This paper carried out a research of the glacier information extraction method based on multi-source remote sensing data, selecting optical imaging, high-resolution image and radar images(SAR) to extract the western Kunlun Mountains Chongce glacier, through the experiments to comprehensive evaluation. First, use a conventional optical remote image glacier extraction methods to experiment and analyse; then, the extraction glacier method using synthetic aperture radar images were studied; as the same time, proposing an extraction glacier method based on image fusion and the corresponding experiments were conducted. In this paper, the main research results summarized as follows:(1) Extraction method based on multi-source remote sensing image fusion is optimal result, which extracts high precision information of glacier and small error as the same time glacial features may also be effective extraction. Through comparing the error of area of the different methods, the results shows that the error ratio of extracting classification after fusioning is significantly reduced, and the extracting results are significantly better than traditional methods. Wherein the minimum error ratio based on Landsat and IRS-P5 fusing is only 0.15%, which is best extraction results.(2) The optimal extraction method is the best extract method for optical image. Using other conventional methods such as ratio method, snow cover index, unsupervised classification and supervised classification method, the extraction effect is obvious- but not universal. The object-oriented multi-scale segmentation extraction method can significantly improve the extraction accuracy, because the consistency area is 99.58 percent, which can better extract feature significant feature types. Using this method to two glaciers extraction results glacier changes are monitoring. The experimental results confirm the glaciers are gradually melting.(3) Extraction results of multi-temporal SAR better than single radar data, but still less than the optical image. Single radar data glacier information extraction is poor and time-consuming pre-processing operations. For multi-temporal radar data using INSAR techniques get on interference analysis, using the extraction method based on texture feature classification to extract the interference result.Through repeated experiments to determine thechoice of mean, variance, homogeneity of the composite image can be divided strongest.Supervise classified extraction results obtained glacier can clearly distinguish mountain glacier and bared land, but can not distinguish between water and glacier region.