Video trajectory predictive tracking using control-based full state observer and filter design

Video target tracking is being a hot research subject applied in solving a wide range of real world electronic monitoring problems. In this work, a predictive tracking approach is presented by adopting control-based full state observer (FSO) tactics. To remedy the tracking precision inadequacy compromised by using only a FSO, a lowpass filter is designed to correct the tracking trajectory as the case may be. The major merit of the proposed approach lies with avoidance of hypothesis for the stochastic noise uncertainties existing in process' dynamics, as otherwise required with Kalman filtering solution. The presented approach is engineered based on an articulate real-world video sequence set, and later tested on four additional real-world video sequence sets. The results suggest that the tracking quality utilizing the proposed approach is not inferior to that using Kalman filter method, and may thus be rendered as an alternative video target tracking technique.