The HyMap hyper-spectral data was used to classify photosyntheticvegetation (PV), non-photosynthetic vegetation (NPV), and exposed soils in a semiaridsavannah environment of McKinlay, northern Queensland, and Australia. Thisstudy aimed to understandhow effective the sub-pixel classificationapproach appliedon hyper-spectral data to distinguish the vegetation and soil features in semi-aridenvironment. In contrast to the per-pixel approach this approach treats the pixelvalue as reflectance sum of its composite features, and shows its componentabundance. The most commonly used sub-pixel classification technique was used inthis research, namely Linear Spectral Unmixing (LSU). End members were used asthe input class, and the result was compared with the standard maximum likelihoodclassification (MLC) using post-classification comparison method The result of thisstudy shows that LSU produced a patchy distribution of classes throughout theimage. The brown soil tends to be over-estimated with respect to other classes. PVfeatures were relatively well-mapped compare to other classes. NPV features haveproblem with domination of exposed soil reflectance. This is equivalent to theprevious studies result that background soil dominates the spectral reflectance inthis environment. According to the qualitative accuracy assessment, LSU hashigher accuracy in representing PV and NPV compare to the traditional MLCclassification.
Copyrights © 2009