Kalman filters applied to target tracking with multiple sensors
Abstract
The objective of this master thesis is to look at an air surveillance system with the ability to integrate observations from multiple radars. A fictional plane is detected by the system, and the goal is to estimate the position and velocity of the plane. To model the movements of the plane, different state space models are suggested. There are alternative measurement vectors, different distributions of the measurement noise and models that take into account bias in the radar measurements.To estimate the position and velocity of the plane, a Kalman filer is used. The various suggestions for state space models are used in the filter. Then the resulting estimates are evaluated by examining results summarized by tables, plots of estimated paths together with true tracks, coverage probability and mean square error. The effect of multiple radars is clearly visible, in particular for an almost linear flight path. However, there are small improvements by introducing velocity in the measurements. Further, that measurement noise with covariance that depends on the distance between radar and plane is not to prefer for a linear flight path. With more realistic paths, the similar covariance is a better alternative than covariance made by a radar simulator.