Hyperspectral Imaging Sensor Based Feature Aided Tracking

We will develop an innovative system called SpecTrac that uses LWIR spectral properties for the tracking of vehicles and dismounts in complex environments. The SpecTrac target models will improve with multiple views and support tracking using features that are invariant to target temperature, viewing geometry, the thermal environment, and the atmospheric conditions. The new approach will allow the tracking of targets that move in close proximity in the presence of occlusion and variable target motion. SpecTrac can also be used to reacquire targets after long time intervals. We will demonstrate the utility of the new spectral algorithms to improve the capability of trackers that use video-rate data acquired from a different sensor. SpecTrac is based on efficient algorithms that will allow real-time tracking on the sensor platform. During Phase I, SpecTrac will be assessed for the long-duration tracking of vehicles and dismounts using a large set of LWIR HSI data and video-rate data acquired from other sources. A detailed commercialization plan is given for the new software.