Knowledge based anomaly detection for ground moving targets

Ground moving target indicator (GMTI) radars are commonly used on airborne platforms in order to detect and track ground targets, such as cars. An operator can use the provided tracks in order to make inference about the behaviour of targets. Prior knowledge, such as road network information, plays a key role in obtaining accurate tracks and thus helping an operator make better inference. This work shows that prior knowledge can also be used for automated anomaly detection when tracking ground targets, thus lowering the workload of an operator. Models for several target behaviours of interest, such as stop-go motion, are presented and tested in simulated examples.