Skip to Main Content
The hyperspectral space consisting of narrow spectral bands is neither an optimal nor an orthogonal feature space when identifying objects. In this paper we consider a means of reducing hyperspectral feature space to a multispectral feature space that is orthogonal and optimal for separation of the objects from background. The motivation for this work is derived from the fact that the retina of the human eye uses only four broad and overlapping spectral response functions and yet it is optimal for detecting objects of multifarious colors in the visible region. In this paper we explore using spectral response functions for the Short Wave Infrared (SWIR) region that are not sharp, but broad and overlapping and even more complex than those found in the retina for the visible region. Treating the measured intensities of the narrow spectral bands as feature vectors of the object of interest, we calculate a new vector space which effectively is a weighted average of the old space, but is optimal for separating the object from the background.