Online maximum likelihood filtering for aircraft tracking under low accuracy observations

Maneuvering aircraft tracking under low accuracy observations is a challenge. To solve this problem, we constructed the corresponding dynamic model for different flying states of aircraft. Then we combined the models of different flying states into single one. Since an accurate dynamic model is a strong constraint for the flying trajectory of aircraft, the trajectory should follow the dynamic characteristic of aircraft regardless of the accuracy of observations. Based on a generalized conditional distribution of target movement on multi-sensor observations, we got an optimal estimation of flying trajectory under the maximum likelihood criterion. By making full use of the association sparseness between the points on the trajectory, we realized an online algorithm. The experiment showed that the proposed algorithm has a significant improvement on localization, prediction and global optimization compared with EKF and DCT filter, while keeping a good real-time performance.