【摘要】 利用Landsat TM遥感光谱信息,构建了多维向量空间下的14个遥感定量指标,采用Fisher逐步判别分析法从这些指标中筛选出8个与沙漠化现状(非沙漠化、潜在沙漠化、沙漠化)密切相关的指标(氧化铁、地表反照率、植被指数)和反映水与植物特性的5个波段(波段1,2,4,5,7),又分辨出6个与沙漠化程度密切相关的变量(氧化铁、地表反照率、植被指数和反映土壤水分和植物水分的2个波段(波段5,7)).利用这些指标建立的判别方程对鄂尔多斯地区沙漠化程度进行了评价,并与2000年的沙漠化程度评价结果对比,发现判别方程的精度达68.9%.结合遥感光谱信息和Fisher分类器可以在遥感图像处理软件下实现沙漠化现状与程度的自动识别,与单纯使用遥感光谱信息进行分类相比,引入遥感代用指标,对于沙漠化现状与程度的分类和分级具有更好的效果.更多还原

【Abstract】 In this paper,under the multi-dimensional vector space,14 quantitive indices of remote sensing were established using the spectrum information of Landsat TM.By adopting the method of stepwise dis- criminant analysis of Fisher,we first screened 6 indices-Fe₂O₃ Albedo,NDVI,and the combination of bands （band 1,2,4,5 and 7） reflecting soil water and vegetation water,which closely correlate with the present condition of sandy desertification （non-sandy desertification land,potential sandy desertification land,and sandy desertification land）;second,we chose Fe₂O₃,Albedo,NDVI,and the combination of bands （band 5 and 7） reflecting soil water and vegetation water as the indices of the sandy desertification.Then,these indices were used to establish a discriminant function for appraising the desertification degree of the study area,and the result was compared with that of 2000,and the precision is 68.9%.The combination of spec- trum information and Fisher classifier can automatically identify the present condition and degree of sandy desertification and,compared with the classifying method of only using spectrum information,the use of the substitute index of remote sensing produce a better effect on classification of the present condition and degree of sandy desertification.更多还原