A kind of modified Kalman filter for visual tracking in capturing noncooperation target aircrafts

To overcome the weakness of the slow convergence of the velocity component in the state vector in the classical Kalman filter (KF), a velocity item, the position difference of the observations, was introduced to serve as the observation velocity to rectify the prediction velocity and accelerate the convergence speed of the state vector, and then improve the tracking instantaneity. A modified mixed KF model, the linear combination of the uniform KF model and the uniform acceleration KF model, was presented to improve the tracking precision, due to the unknown of the real motion model of the noncooperation target aircraft and the weakness of the classical KF motion model' unchangeableness. Besides, the minimum mean square error (MSE) between the prediction points with the reference points, accompanied by the system error covariance, was designed to visually evaluate the performance of the modified KF. The simulation results indicate that the modified mixed KF with the observation velocity could effectively reduce the tracking error of the noncooperation target aircraft, and improve the tracking precision and instantaneity.