Hyperspectral Identification for Collaborative Tracking

We will develop a system called HyperTrack that will significantly advance the capability of target tracking systems. HyperTrack is based on new algorithms for acquiring target models and new methods for using these models for tracking. An innovative physics-based algorithm is used to acquire a distinctive spectral model from one or more images of an unknown target. An associated algorithm is used for the acquisition of a distinctive target geometric model. Detailed physical models will be used to allow the target spectral and geometric models to adapt to changing environmental conditions while at the same time adapting to maximize contrast with the background. A general Bayesian technique will be developed to combine information that is acquired from different sensors on different platforms. The new model acquisition and tracking algorithms will be assessed using a range of airborne imagery that includes vehicle and dismount targets. The algorithms will be derived to allow efficient operation on modest hardware. We will also consider the transition of the HyperTrack software for a range of applications.