Adaptive tracking system for a manoeuvring target using images with correlated noises

An effective scheme is proposed for estimating the state parameters of a manoeuvring target from a noisy image sequence, providing the sequence contains correlated noises and the trajectory of the target is disturbed by an unknown acceleration. For the correlated noises, a first-order difference operator is applied to the original image sequence to generate an artificial measurement sequence with only uncorrelated white noises. For the unknown acceleration, a Kalman filter augmented by a semi-Markov process and the bayesian theory is applied to form an adaptive filter. In the proposed tracking system, the filter first generates an artificial measurement at each sampling time from the observation. It then utilizes an artificial measurement sequence up to the current time instance to predict the a posteriori probability of each possible acceleration state. The weighted average of acceleration, where the weight is the a posteriori probability, is applied to derive the optimal estimates of the state parameters. Several computer simulations with external force applied at unknown times are performed to demonstrate the applicability and superiority of the proposed system.