【摘要】 西北黄土高原区面积广,滑坡数量众多,利用遥感技术加快减灾、防灾,开展编目、区划研究,具有重要的研究意义。目前,利用人机交互解译、计算机分类,滑坡灾害解译和识别已取得许多成功的经验。受数据源、数据处理和解译等交叉学科发展的影响,应用多侧重影像预处理、解译标志建立和技术方法研究,滑坡灾害遥感解译不确定性等关键问题研究较少,解译标志也过分关注色和形描述,与滑坡发育规律密切的地形地貌等重要解译标志尚需完善。本文以西北黄土高原陕西延安、宁夏彭阳等地为研究区,从尺度效应、波段组合、数据融合等方面分析滑坡灾害遥感解译的不确定性,结果表明地貌、滑坡和承灾体、岩土体变化呈等级结构,与之对应遥感数据的时空尺度不同,利用ETM+,Spot 5和QuickBird等不同分辨率数据进行识别分类,能减少解译的不确定性;波段组合能有效反映地层岩性时,更适合滑坡识别和分类;高空间分辨率数据的融合算法以主成分或HIS较佳。基于分析结果,采用Spot 5经自然彩色变换主成分融合,依据研究区黄土滑坡的典型特征及影像表现,利用人机交互解译技术建立滑坡解译标志,为编目和区划、滑坡灾害计算机解译等提供依据。更多还原

【Abstract】 The northwestern Loess Plateau covers vast area,and numerous landslides occurred in the area.It is important to use remote sensing technology for inventories and zoning studies as well as effective mitigation of the disasters. Currently,landslides interpretation has been used widely,as it takes advantage of visual,human-computer interaction as well as computer classification.However,there are few studies on scale effects of remote sensing landslides,and uncertainties in interpretation,restricted largely by development of data source and interdisciplinary nature.Interpretation keys have been excessively concerned with the color and shape descriptions,while the important interpretation keys in temporal and spatial variations still need to be improved,such as the topography that is closely associated with the landslide development.This paper takes the northwestern Loess Plateau as a case study,based on the uncertainties analysis of landslides hazard interpretation of remote sensing image.The results show that landscape,landslides and hazard-affected bodies,rock and soil changes in the northwestern Loess Plateau have a hierarchical structure,corresponding to different time and space scales of the remote sensing data.Loess landforms,landslides,hazard-affected body should be used.ETM+,Spot 5,and QuickBird data were used for identification;visual interpretation in the same band combination are more suitable for regional landslide inventory and zoning.The application of high resolution satellite data such as QuickBird have more advantages and can identify more information than low spatial resolution data.When band combination can effectively reflect the lithology,they are more suitable for identification and classification of landslides.High spatial resolution data fusion algorithms are better for principal components analysis or HIS.Using Spot 5-transformed natural color and merging with the principal component analysis and combining the regional typical loess landslides and video performance, this study tried to establish the landslide interpretation keys in the northwestern Loess Plateau by human-computer interactive interpretation,which will be useful for the landslides identification and classification,and the inventory and zoning study.更多还原