摘要
以位于黑河流域北大河下游的金塔绿洲为例,以LandSat TM为数据源,在确定绿洲NDVI(normalizeddifference vegetation,INDV)阈值,对影像进行主成分分析并计算基于灰度共生矩阵的纹理指数的基础上,运用决策树分类法,进行了天然绿洲和人工绿洲的区分,并将区分结果与单纯使用INDV和最大似然分类结果进行了比较.结果表明,结合纹理与植被指数区分天然绿洲与人工绿洲的总精度为62.50%,Kappa系数k=0.416;与只用INDV区分的方法相比,总精度提高1.50%,k提高2.60%;与最大似然法相比,总分类精度提高0.90%,k提高2.20%.说明将纹理特征与植被指数相结合的方法可以在一定程度上提高天然绿洲与人工绿洲的区分精度.
Taking Jinta oasis in downstream of the North River which is located in Heihe River as an example, and using the spectrum TM image as data sources, NDVI （Normalized Difference Vegetation Index） threshold was confirmed, principal component for image was analyzed and texture features based on gray-level co-occurrence matrix was calculated to distinguish natural and artificial oasis＇ division. Obtained results were further compared with the results only from using NDVI threshold and from the maximum likelihood method. Total precision of distinguish natural and artificial oases is 62. 50% and Kappa coefficient is 0.416 by the method of combining texture and vegetation index together. Total classification accuracy increased about 1.50% and Kappa coefficient increased about 2.60% compared with the results from the method only using NDVI threshold. Compared with maximum likelihood＇ s method, total classification accuracy and Kappa coefficient increased 0. 90% and 2. 20%, respectively. As a result, combination of the texture characteristics and vegetation index can increase the accuracy of distinguish artificial and natural oases.
出处
《宁夏大学学报（自然科学版）》
CAS
2013年第1期88-92,共5页
Journal of Ningxia University(Natural Science Edition)
基金
国家重点基础研究发展计划基金资助项目(2009CB421306)
国家大学生创新性实验计划基金资助项目(101073027)
