Jump-diffusion based sampling algorithm for target tracking and recognition

A new random sampling algorithm for recognition and tracking of an unknown number of targets and target types is presented. Taking a Bayesian approach we define a posterior measure on the parameter space by combining the observed data likelihood with a prior based on airplane dynamics. The Newtonian force equations governing the airplane motion are utilized to form the prior density on the airplane positions. The sampling algorithm based on Jump-diffusion processes, first introduced by Grenander and Miller (1991), is derived for generating high probability estimates of target positions, orientations and types from the posterior measure. Results are presented from its joint implementation on the Silicon Graphics workstation and the DECmpp SIMD machine distributing data-simulation, visualization and computation over network