Hyperspectral Persistent Surveillance Exploitation Algorithms

We will develop the INV-PS system for combining motion constraints, spectral properties, and spatial properties for target detection and tracking in persistent surveillance data. INV-PS will learn adaptive spectral and spatial models that can be applied to targets with a wide range of characteristics. The models will support detection and tracking in complex environments over time intervals that vary from less than a second to several days. The approach is derived to allow tracking in the presence of illumination changes, spectral mixing, occlusion, and variable target motion. In addition, INV-PS is designed to allow different sources of information to be incorporated in a principled manner. INV-PS uses efficient algorithms that support real-time processing on the sensor platform. The new models and algorithms will be assessed over a range of data. A detailed commercialization plan is given for the new software.