Classification of TiungSAT-1 MSEIS Data for Land Cover Mapping with Hyperspectral Analysis Approach

Abstract

The low to medium spatial resolution satellite data can still be utilized to meet the requirement for a certain level of land cover mapping. However, the land cover classifications of assigning pixel-by-pixel basic to specific land cover classes have been known to be a problematic phenomenon that limits the accuracy of classification. This paper examines the result of utilizing the hyperspectral approach as a recent alternative solution to the above problem to investigate whether or not the sensitivity allowed in the latter can increase the classification accuracy. Using TiungSAT-1 MSEIS data as input, comparative analysis were also performed with classical Maximum Likelihood classification. The results of this study clearly indicate that hyperspectral analysis can improve classification accuracy of multi-spectral data, especially when the mixed pixels are of great concern.