Spectral Screened Orthogonal Subspace Projection for Target Detection in Hyperspectral Imagery

We introduce several new approaches in increasing spectral screening accuracy and reliability and study their applicability for orthogonal subspace projection-based target detection. Spectral screening is a technique used for selecting a representative subset of spectra from hyperspectral data. The subset is formed such that any two spectra in it are dissimilar, and for any spectrum in the original image cube, there is a similar spectrum in the subset. Spectral screening is performed in a sequential manner; at each step, the subset is increased with a spectrum dissimilar from all the spectra already selected. The procedure, using the spectral angle as similarity measure, is employed in a variety of algorithms for linear unmixing and data compression. We modified this algorithm such that at the selection step the spectrum with the largest distance from the set is selected. While not introducing additional computational complexity, the maximum spectral screening (Max SS) algorithm ensures that the overlap among the representatives is minimized, and that the solution is deterministically obtained, eliminating the need for multiple experiments. We also investigated the alternative approach in which at each step the spectrum with the smallest distance (but larger than a threshold value) is selected. In a distinct direction, we also studied the efficiency of the two algorithms by designing detection filters obtained as the classification projector matrices based on the spectral subset using regular or kernel-based orthogonal subspace projections (KOSP and OSP, respectively). The developed algorithms were tested on Hyperspectral Digital Collection Experiment (HYDICE) hyper-spectral data using the spectral angle and the spectral information divergence. The results indicated that Max SS outperforms regular spectral screening detection and minimum spectral screening (Min SS).