Spectral matching based on hidden Markov model

Combined with traditional image information and spectral information, hyperspectral remote sensing could not only get the space information about the surface of the earth, but also obtain continuous spectrum of single pixel. Spectral matching technique is one of the key technologies of imaging spectroscopy remote sensing classification and target detection. Spectral characteristics can be used to identify surface features category in hyperspectral remote sensing. The traditional method of spectral matching includes the minimum Euclidean distance matching, spectral angle matching and spectral similarity matching. SAM (spectral angle matching) is better than others, but the discrimination is not high, and usually could not get a satisfactory result. This paper gives a proposal that introducing and using the hidden Markov model to describe the pixel spectral characteristics, and then compare this method with several commonly used methods by using the standard USGS spectral library data in the experiment.