Enhanced lowpass filter based video predictive tracking for target with high mobility

A lowpass filter is designed to perform the predictive tracking of highly mobile target. The proposed filter consists of linear term and inertial terms, with each respectively responsible for non-high mobility and high mobility. The target high-mobility is detected by measuring the velocity change and the inertial term is to be reconfigured to dominate the filtering process when the target moves in a highly mobile fashion. The tracking tests are conducted on the given set of video sequences and compared against the predictive tracking performed by Kalman filter, suggesting the much better tracking quality and feasibility of the proposed approach.