Abstract:
Kalman filter-based solutions proposed for nonlinear systems are frequently used in bearing-only tracking applications. Due to the physical conditions of these tracking applications, the measurements gathered may contain a high amount of noise. For example, if the measurement sensors are too far from the target being tracked, a small error in the calculated bearing or a small amount of noise exposure will cause the uncertainty in the tracking system to increase significantly. Since the effect of this large amount of noise can only be eliminated to a certain extent by Kalman filter-based solutions, tracking performance may decrease in these applications. In this thesis, various statistical and machine learning-based noise removal methods will be applied to reduce the noise in bearing measurements obtained with sensors. Then, these noise-reduced measurements will be used in Kalman filter-based solutions in bearingonly tracking. The effects of noise reduction methods on tracking performance will be compared with simulations on real vessel trajectories.
