摘要
针对绿洲土地覆盖/利用特点,建立了反映绿洲荒漠化类型及特点的分类系统。对原始SPOT数据进行变换,提取NDVI、K-L、NDMI等14个特征波段,并利用波段的类间可分性进行了参与分类的最优波段选取。根据地物的波谱特性和空间特性建立了决策树模型,对研究区荒漠化信息进行了提取。分类结果表明,风蚀荒漠化及盐碱化是吐鲁番主要的荒漠化类型。其中,中度风蚀荒漠化土地分布在绿洲边缘及外围,2004年占研究区总面积的64.05%;盐碱化土地及轻度、重度风蚀荒漠化土地分布在绿洲内部,三者分别占研究区总面积的6.60%、7.58%和2.23%,吐鲁番盐碱化程度高于托克逊县,风湿荒漠化程度相反。本文提出的基于决策树分类的绿洲荒漠化信息提取方法可以充分利用地物的光谱信息与地理位置信息,较好的区分了混淆地物,达到了较好的分类效果。
Oasis is the major portion with relatively high productivity and hosts a lot of lives in arid areas. It takes up only 4% of the area but sustains 95% of the population and above 90% of social wealth in Xinjiang, China, which is a typical arid area in the world. Oasis is the citadel to explore in desert and mountainous areas, and therefore plays a fundamental role in the development of society and economy in Xinjiang. The Turpan Basin is a typical oasis-desert interlaced area in eastern Xinjiang, and an important part of the ecologically and environmentally fragile zone of northwestern Oasis in China. Desertification is the most serious environmental problem in the Turpan Basin. It is important to accurately and rapidly monitor desertification. Satellite remote sensing has become a commonly used technique in desertification monitoring. High resolution remote sensing imagery, SPOT data, was used in this study. First, a classification system which is able to reflect desertification types and characteristics was built in terms of land use/cover characteristics. A series of atmospheric and geometric corrections and imagery enhancement were performed on the SPOT imagery before using it. There were 10 characteristic bands derived from the SPOT data, such as NDVI K-L, and NDMI. Inter-class separability was applied to choose the optimum combination of bands. Based on spectrum and spatial characteristics of different features, combined with supervised classification results, a decision tree was built to derive land use/cover information, in particular the desertification distribution. Results showed that wind erosion desertification and salinization were the major desertification types in the study area. Moderate wind erosion desertification was distributed on the edge and the periphery of oasis, accounting for 64.05% of the total area. Salinization land, mild and serious wind erosion desertification were distributed within the oasis, accounting for 6.6, 7.58 and 2.23% of the total area, respectively. The salinization degree of Turpan was higher than that of tuokexun, whereas wind erosion desertification degree was contrary to the former. The decision tree classification method made full use of spectral and spatial information of land features, and seems capable of discriminating land features.
出处
《资源科学》
CSSCI
CSCD
北大核心
2011年第6期1204-1210,共7页
Resources Science
基金
秦皇岛市科学技术研究与发展计划(编号:201001A429)
