Parallel implementation of algorithms for standoff detection in hyperspectral imagery

Hyperspectral imaging systems, used in conjunction with appropriate detection and recognition algorithms, have demonstrated to be very appropriate tools for standoff detection in many different environments. Compared to other techniques available such as multispectral imaging, which typically collects only tens of images, hyperspectral instruments are capable of collecting hundreds of images, corresponding to different wavelength channels, for the same area on the surface of the Earth. While developments in hyperspectral technology hold great promise for advanced standoff detection, they create new processing challenges. In particular, the price paid for the wealth spatial and spectral information available from hyperspectral sensors is the enormous amounts of data that they generate. However, several applications exist where having the desired information calculated in real-time or near real-time is desired, e.g. those focused on the detection and tracking of forest fires, oil spills and other types of chemical contamination. In this paper, we discuss parallel implementations of a novel morphological algorithm for standoff detection in hyperspectral imagery. Several illustrative processing examples are provided, thus allowing an evaluation of the speedup factors provided by implementing the above algorithm, characterized by its combined use of spatial and spectral information, on massively parallel computer facilities.