Self-adaptive Constant Acceleration Model and Its Tracking Algorithm Based on STF

Abstract

In sensor target tracking and data processing, building a dynamic model of target motion plays an important role. Current statistical (CS) model is one of better target dynamic models that were widely applied in practical problems now. However, for tracking nonmaneuvering targets, using CS model will cause large error. Furthermore, the conventional tracking algorithm corresponding to CS model is based on Kalman filter (KF) or extended Kalman filter (EKF). But KF and EKF have bad robustness on the modeling uncertainty, and are sensitive to the initial conditions. In order to overcome the shortcomings of CS model and its tracking algorithm, a self-adaptive constant acceleration (CA) model and its tracking algorithm (ACA-STF) is presented by comparison and study of CS model and CA model and introducing a fading factor of strong tracking filter (STF) in the paper. The algorithm can self-adaptively adjust the covariance matrix of process noise and tune a filtering gain matrix on line. The theoretic analyses and simulation results show that this algorithm has better tracking performance to track non-maneuvering targets and maneuvering targets than CS model and its tracking algorithm.