Study On Topographical Objects Classification Method Based On Hyperion Data

There is a very important demand at present in automatic classification of topographical objects by utilizing the remotely sensed image . With the increasing demand and high requirement of society to the special image ,the special made from low resolution image can not satisfy the requirements in many fields .Hyperspectral remote sensing data has the characteristic of combining both image data and spectral data into one data set, each pixel of a hyperspectral image an entire spectrum is present. Compared with conventional multi-spectral remote sensing data, hyperspectral remote sensing data can be more useful for the identification and classification of surface Materials.Topographical objects classification analyses were performed at Huxi and Fuyuzhan forestry centre at the boundary of Heilongjang province and Jilin province using EO-1. Classification analyses based on Linear Spectral Mixture Model were performed according to the characteristics that Hyperion has vast data and many bands .Traditional supervised classification method (Minimum Distance and Maximum Likelihood) and unsupervised classification method (K-Means and ISODATA) were also used in this paper. The conclusions are as follows by comparing the classification results: The classification method based on Spectral Mixture Model is more better than traditional classification method and the supervised classification is better than the unsupervised classification. The reason is that The classification method based on Spectral Mixture Model using the advanced photic elements and change the object of classify from pixel to sub-pixel.But the classification method based on Spectral Mixture Model also has the problem of it's difficult to get the endmember , so should use the classification method combine the traditional classification method or to get the endmember with the traditional classification method , and to provide the good method for application .