GMTI Radar Target Classification Using Spectral and Tracking Data

TSC will mature the Human, Vehicle, Animal and Clutter (HVAC) discrimination algorithms developed in Phase I for use in GMTI radars including the airborne foliage penetration (FOPEN) UHF airborne surveillance systems and tower-mounted radars under development by the US Army CERDEC I2WD. In Phase I, a limited set of FORESTER data was used to show the feasibility of generating multiple-CPI features and successfully classifying humans versus animal targets. In Phase II, TSC will continue focusing on developing multiple-CPI processing techniques and investigate methods of estimating the accuracy of classification decisions. To support algorithm development and testing, TSC will collect UHF radar signatures of dismounts and commonly occurring animals performing natural activities using its existing equipment. This collection will include simultaneous video recording and calibration to enable absolute RCS measurements. TSC will also investigate the extraction of track-based features and group structure and size information to aid in target discrimination. TSC will additionally explore the use of Deep Neural Networks to perform HVAC discrimination and group size and structure estimation. TSC will deliver the collected radar database and conduct a demonstration of dismount versus animal classification for the Army. The algorithms will also be adapted for X and Ku-band sensors.