基于多源遥感影像数据的冰川信息提取方法研究

发布时间：2018-01-08 15:10

本文关键词：基于多源遥感影像数据的冰川信息提取方法研究　出处：《兰州交通大学》2016年硕士论文　论文类型：学位论文

【摘要】：冰川信息的提取是研究冰川消融、气候变化等方面的全球化热点问题。遥感技术已经成为冰川提取中的必要技术手段。遥感事业的迅猛发展形成了遥感数据多源化的特点。研究基于多源遥感数据的冰川提取方法可以对现有的多源遥感数据进行综合利用,提高冰川提取的精度。本文开展了基于多源遥感数据的冰川信息提取方法研究,选用光学影像,高分辨率影像以及雷达影像(SAR)应用于西昆仑山脉的崇测冰川信息提取中,通过实验进行综合评价。首先选用传统的光学遥感冰川提取方法进行了实验和分析;之后对合成孔径雷达影像的冰川提取方法进行了研究;提出了基于图像融合的冰川提取方法并进行了相应实验。本文主要的研究成果概括如下:(1)基于多源遥感影像融合的提取方法最优,提取结果精度高误差小,并且可以有效提取冰川特征。通过对比各方法的面积误差比,结果表明先融合再分类提取的误差明显降低,且提取结果显著优于传统提取方法。其中基于Landsat与IRS-P5融合的误差比最小仅为0.15%,提取结果最优。(2)面向对象法是光学影像的最优提取方法。比值法、雪盖指数法、非监督分类与监督分类法,提取效果明显但都不具有普适性。面向对象多尺度分割提取方法能明显提高提取精度,面积一致性达99.58%,能较好地提取具有显著特征类型的地物。利用该方法对两期冰川提取结果进行了冰川变化监测研究,结果证实了本实验区冰川正在逐步消融。(3)多时相SAR的提取结果优于单时像SAR但不及光学影像。对SAR影像的提取研究发现:单幅雷达数据冰川信息提取效果较差且预处理操作耗时长。对多时相雷达数据采用INSAR技术进行干涉分析,对干涉结果采用基于纹理特征的分类方法进行提取,通过多次反复实验确定选用均值、方差、同质性合成影像可分性最强。进行监督分类得到冰川提取结果能明显区分山体裸体与冰川区,但不能区分水体与冰川区。
[Abstract]:The extraction of glacier information is to study the melting of glaciers. The technology of remote sensing has become an essential technique in the extraction of glaciers. The rapid development of remote sensing has formed the characteristics of multi-source remote sensing data. The research is based on multi-source remote sensing data. The existing multi-source remote sensing data can be comprehensively utilized by the glacier extraction method. In order to improve the accuracy of glacier extraction, the method of extracting glacier information based on multi-source remote sensing data is studied in this paper, and optical image is selected. High resolution images and radar images are used to extract the information of the glacier survey in the West Kunlun Mountains. Firstly, the traditional optical remote sensing glacier extraction method is used to carry out the experiment and analysis. Secondly, the glacier extraction method of synthetic aperture radar (SAR) image is studied. A glacier extraction method based on image fusion is proposed and the corresponding experiments are carried out. The main research results of this paper are summarized as follows: 1) the extraction method based on multi-source remote sensing image fusion is optimal. By comparing the area error ratio of each method, the result shows that the error of first fusion and then classification is obviously reduced. The error ratio based on the fusion of Landsat and IRS-P5 is only 0.15%. The object oriented method is the best extraction method for optical image. Ratio method, snow cover index method, unsupervised classification and supervised classification method. The extraction effect is obvious but not universal. Object-Oriented multi-scale segmentation and extraction method can significantly improve the accuracy of the extraction area consistency up to 99.58%. The method is used to study the glacier change monitoring of the two periods of glacier extraction results. The results confirm that the glaciers in this experimental area are gradually melting. The extraction result of multitemporal SAR is better than that of monochronous SAR but less than that of optical image. It is found that the extraction of SAR image is better than that of optical image. The extraction effect of glacier information from single radar data is poor and the pretreatment time is long. The interference analysis of multi-phase radar data is carried out by using INSAR technology. The interference results are extracted by texture feature based classification method, and the mean value and variance are determined by repeated experiments. The results of glacial extraction can distinguish the naked mountain body from the glacier area, but it can not distinguish the water body from the glacial region.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P237;P343.6

【参考文献】