Group state estimation algorithm using Foliage Penetration GMTI radar detections

This paper describes an algorithm for integrating detections from Foliage Penetrating (FOPEN) Ground Moving Target Indicator (GMTI) radar, to recognize groups of dismounts moving through dense foliage, and to estimate the group states, including the group sizes and the directions of their movements. Difficulties of processing FOPEN GMTI radar detections are best characterized as low target-state-dependent detection probabilities and high non-uniform persistent false alarm densities. To overcome these difficulties, we use the Sum-of-Gaussian (SOG) or Gaussian-Mixture (GM) Cardinalized Probability Hypothesis Density (CPHD) method to detect and track individual dismounts, and then, apply a group dynamics recognition method to the CPHD outputs to recognize the formation and the behavior of the dismounts groups.