基于RS与GIS的冰川变化参数信息提取研究

发布时间：2018-03-24 06:02

本文选题：传统遥感解译方法　切入点：面向对象信息提取方法　出处：《西北师范大学》2013年硕士论文

【摘要】：冰川是气候变化的指示器和储存体，其变化是全球变化的重要指标。随着遥感技术的发展，借助遥感手段研究冰川的性质和特征、监测冰川的动态变化是冰川学研究的重要手段之一，为全球冰川动态变化研究提供了及时、准确、廉价的数据源。本论文的研究内容主要分为三个部分：应用遥感影像提取冰川信息的传统方法和面向对象的信息提取方法比较；在天山台兰河流域探讨面向对象信息提取方法在冰川自动化提取冰川面积信息的优势；应用地形图DEM和SRTM4求取冰川区的相对厚度变化。受数据限制，研究选取了我国大陆型冰川分布的三个典型冰川区：阿尔泰山喀纳斯河源区，天山台兰河流域，祁连山黑河葫芦沟小流域地区。在阿尔泰山喀纳斯河源区综合目视解译以及计算机自动解译的方法，将冰川解译的传统方法及面向对象图像特征提取方法应用于冰川边界提取中，比较各种方法的优劣；冰储量是表征冰川水资源变化状况的最直接参数，冰储量估算的最核心问题是冰川厚度的测量，对黑河葫芦沟小流域近40年来的厚度退缩进行计算，估算黑河流域冰川变化情况。本论文的主要研究成果如下：（1）通过将传统遥感解译的方法以及面向对象的解译方法应用到喀纳斯河源区，，研究发现：目视解译精度较高，虽然耗时耗力，但适用于任何类型的冰川解译。传统计算机自动解译方法各有优缺，阈值法操作简单，但是区分效果不明显；雪盖指数法比直接用阈值法要精确些，结果也不尽人意；非监督分类方法精度较低；监督分类方法准确度明显高于非监督分类，但是部分积雪和岩石还是容易被错分；波段比值法操作相对简单，精度较高。决策树自动生成研究避免了人工生成决策树时的主观性，该方法无法识别样点数据中的误差，有明显的错分；面向对象的提取方法能更好的提取各种具有特征类型的地物，应用于冰川的提取具有较高精度。（2）选取天山台兰河流域这一典型冰川区应用面向对象的方法进行冰川的自动化提取。得到：2011年，通过面向对象影像解译得到台兰河流域共发育现代冰川109条，冰川总面积385.38km~2，冰川平均面积为3.54km~2。近40年来台兰河流域冰川总面积共退缩50.06km~2，平均每年退缩0.94km~2，冰川平均面积减小0.31km~2，表明该地区冰川处于强烈退缩状态。研究发现面向对象的提取方法能高效提取冰川，且能有效抑制云和雪的干扰，在自动解译冰川方面有良好的应用前景。（3）多源DEM数据的出现，使获得冰川厚度变化成为可能。利用地形图DEM、SRTM-DEM数据研究冰川厚度变化，从二维转向三维更为精确的研究冰川变化。对黑河葫芦沟小流域1956-2000年冰川相对厚度变化进行计算发现冰川厚度平均减薄1.5m，最大减薄15m，最大积累3m。
[Abstract]:Glaciers are indicators and reservoirs of climate change and their changes are important indicators of global change. With the development of remote sensing technology, the nature and characteristics of glaciers are studied by means of remote sensing, Monitoring the dynamic changes of glaciers is one of the important means of glaciology research, which provides a timely, accurate and cheap data source for the study of global glacier dynamic change. The research content of this paper is divided into three parts: comparing the traditional method of extracting glacier information from remote sensing image with the method of object oriented information extraction; This paper discusses the advantages of object-oriented information extraction method in automatic extraction of glacier area information in Tianshan Tailan River Basin, and uses topographic map DEM and SRTM4 to calculate the relative thickness variation of glacier area, which is limited by data. Three typical glacial regions of continental glacier distribution in China are selected: the Kanas River Source area in Altai Mountain, the Tailan River Basin in Tianshan Mountain, and the Tailan River Basin in Tianshan Mountain. Hulugou small Watershed area of the Heihe River, Qilian Mountains. Comprehensive visual interpretation and computer automatic interpretation in Kanas River source area, Altai Mountain, The traditional glacier interpretation method and object oriented image feature extraction method are applied to the glacier boundary extraction to compare the advantages and disadvantages of various methods, and ice reserves are the most direct parameters to represent the change of glacier water resources. The core problem of ice reserve estimation is the measurement of glacier thickness. The thickness retreat in Hulu Valley of Heihe River in recent 40 years is calculated to estimate the glacier variation in Heihe River Basin. The main research results of this thesis are as follows:. 1) by applying the traditional remote sensing interpretation method and the object-oriented interpretation method to the Kanas River source area, it is found that the visual interpretation accuracy is high, although it is time-consuming and labor-intensive. But it is suitable for any type of glacier interpretation. The traditional computer automatic interpretation method has its own advantages and disadvantages, the threshold method is simple, but the discrimination effect is not obvious, the snow cover index method is more accurate than the direct threshold method, and the result is not satisfactory. The accuracy of the supervised classification method is obviously higher than that of the unsupervised classification method, but some snow and rock are still easily misclassified, and the band ratio method is relatively simple. The research of automatic generation of decision tree avoids the subjectivity of artificial generation of decision tree, and the method can not identify the error in the sample data, and there are obvious errors. Object-Oriented extraction method can better extract all kinds of feature objects, and it has high accuracy in glacier extraction. 2) selecting the typical glacial region of Tianshan Tailan River Basin to carry out automatic glacier extraction by using object-oriented method. The following conclusions are obtained: in 2011, 109 modern glaciers have been developed in the Tailan River Basin by object oriented image interpretation. The total glacial area is 385.38 km2, and the average glacier area is 3.54 km2. In the last 40 years, the total glacial area in Tailan River basin has receded by 50.06 km / 2, averagely receding 0.94 km2 / year, and the average glacial area has decreased by 0.31 kmm2, indicating that the glacier in this area is in a state of intense retreat. The method of object extraction can extract glaciers efficiently. It can effectively suppress the interference of cloud and snow, and has a good application prospect in the automatic interpretation of glaciers. 3) the appearance of multi-source DEM data makes it possible to obtain glacier thickness variation. The relative thickness of glacier in Hulugou small watershed of Heihe River was calculated from 1956 to 2000. The results show that the average thickness of glacier is 1.5 m, the maximum thickness is 15 m, and the maximum accumulation is 3 m.
【学位授予单位】：西北师范大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P343.6;P208;P237

【参考文献】