Ground moving target tracking by airborne radar provides situational awareness of vehicle movements in the supervised region. Vehicles are detected by applying space time adaptive processing to the received multi channel radar data. The detections are then fed to a tracking algorithm that processes them to tracks.
In literature, radar signal processing and ground target tracking are treated as two separate topics and results are not validated by experimental data. The first objective of this thesis is to provide a closer link between these fields. The second objective is to show that tracking performance can be improved by providing additional data from the radar signal processing to the tracking step. The third objective is to validate the algorithm and the performance improvement using experimental data.
As a result this thesis presents a unified treatment of ground moving target tracking from radar raw data to established tracks. A complete reference algorithm for ground moving target tracking based on the Gaussian mixture probability hypothesis density filter is presented. In particular, Jacobians of the observation process are derived. They are presented in such a form that immediate implementation in a programming language is possible.
In the course of this thesis a measurement campaign with the experimental radar PAMIR of Fraunhofer FHR was conducted. The experiment included two GPS equipped reference vehicles and a multitude of targets of opportunity. Tracking results obtained with this experimental data and the reference tracking algorithm of this thesis are shown.
The thesis also enhances the reference target tracking algorithm by a parameter that characterizes the variance of the direction of arrival measurement of the target signal. This parameter is determined adaptively depending on the estimated signal strength and the clutter background. The major contribution with regard to this enhancement is a thorough experimental validation: Firstly, a comparison between GPS based measurements and radar based measurements of the direction of arrival shows that this variance captures the distribution of measurement errors excellently. Secondly, tracking results are compared to the GPS tracks of the ground truth vehicles. It is found that the enhanced algorithm yields superior track quality with respect to both track accuracy and track continuity.