利用高光谱数据进行地物识别分类研究

Using Hyperspectral Data for Land Cover Identification and Classification

分析了传统统计分类方法在高光谱影像地物分类中的弊端,提出并详细讨论了基于端元的监督分类技术.利用端元监督分类技术对LASIS高光谱影像进行分类,同时应用IsoData非监督分类技术即自动迭代聚类对高光谱影像进行分类.分析比较了两种分类结果,表明基于端元的监督分类技术更能满足对地物识别分类的需要.

The limitation in hyperspectral land cover classification is discussed, when using traditional statistical classification methods. An endmember-based supervised classification approach is proposed and discussed. The experimental data was acquired by imaging spectrometer （LASIS）. The endmember-based supervised classification procedure was used for the hyperspectral image. The IsoData unsupervised classification procedure that automatically iteratively clusters the pixels was also used. The results of supervised and unsupervised classification are compared. It shows that the endmember-based supervised classification for land cover is much more satisfied.