A New Hybrid Strategy Combining Semisupervised Classification and Unmixing of Hyperspectral Data

Spectral unmixing and classification have been widely used in the recent literature to analyze remotely sensed hyperspectral data. However, few strategies have combined these two approaches in the analysis. In this work, we propose a new hybrid strategy for semisupervised classification of hyperspectral data which exploits both spectral unmixing and classification in a synergetic fashion. During the process, the most informative unlabeled samples are automatically selected from the pool of candidates, thus reducing the computational cost of the process by including only the most informative unlabeled samples. Our approach integrates a well-established discriminative probabilistic classifier-the multinomial logistic regression (MLR) with different spectral unmixing chains, thus bridging the gap between spectral unmixing and classification and exploiting them together for the analysis of hyperspectral data. The effectiveness of the proposed method is evaluated using two real hyperspectral data sets, collected by the NASA Jet Propulsion Laboratory's airborne visible infrared imaging spectrometer (AVIRIS) over the Indian Pines region, Indiana, and by the reflective optics spectrographic imaging system (ROSIS) over the University of Pavia, Italy.