Evaluation of Unscented Kalman Filter and Extended Kalman Filter for Radar Tracking Data Filtering

This paper focuses on the issue of nonlinear data filtering in radar tracking. Through the analysis on the extended Kalman filter (EKF) and the unscented Kalman filter (UKF), which are both nonlinear filters, we find that the accuracy of the extended Kalman filtered data image was not ideal for radar tracking data filtering, while UKF can achieve better performance. The evidences show that, while comparing with curves dealt with EKF, the curves obtained by UKF in the situation of radar tracking is able to get more accurate results because the mean and variance of the nonlinear function can be estimated more accurately by means of unscented transformation, and the computation complexity is reduced significantly by avoiding to calculate the Jacobian matrix.