Abstract:

A Kalman filter with online parameter adjustment functionality and its application in ship target tracking is proposed.In practical systems,noise statistics has dynamic features.Gennerally, in conventional Kalman filtering, the variances of state and measurement noise are assumed to be fixed and known beforehand. However, this assumption does not hold in practice generally.The Kalman filter without correct information on noise covariances could produce poor performance on filtering.In the worst case,it leads the filter to diverge.In order to overcome this problem, a method which adaptivly adjusts the noise mean and covariance to track the dynamic noise is proposed,so as to improve the precision of target tracking.Online realization can effectively use part of data to update iteration, which  reduces calculation and is easy for engineering realization .Finally, the results of the ship target simulation demonstrate the effectiveness of the porposed algorithm.

Key words: online realization；Kalman filtering;target tracking