STUDY OF PARTICLE AND KALMAN FILTER FOR IMAGE TRACKING

Abstract

An unscented Kalman filter (UKF) has been extensively used for tracking problems. UKF has been used earlier to generate proposal distributions to turn a generic particle filter to a high-performance unscented Particle filter (UPF). This thesis uses the unscented Kalman filter to generate sophisticated proposal distributions that seamlessly integrate the current observation, thus greatly improving the tracking performance. The Unscented Particle Filter, those that the reduce the size of the space explored by the particles or those that perform a local exploration of the likelihood surface before predicting new particles. The use of the generic particle filter (PF) algorithm is well known for target tracking, but it can not overcome degeneracy of particles and cumulation of estimation errors. In this paper, we propose an improved PF algorithm called PF-RBF. Radial- basis function network (RBFN) is used in the sampling step for dynamically constructing the process model from observations ...