The impact of band selection on hyperspectral point target detection algorithms

In this paper, we explore the influence of band selection and dimensionality reduction of hyperspectral data on three point target detection algorithms. We wish to reduce the computational burden and to maximize the algorithms' performance by taking into consideration high spectral correlation. In order to measure the discrimination capability of target detection algorithms, we implemented a metric to quantitatively evaluate our algorithm for a particular combination of target signature, spectral cube, and bands chosen. Band selection was done in several ways; we evaluate our results both with exhaustive search and a "sub-optimal" selection algorithm.