Skip to Main Content
Detecting the activities and predicting the tendencies of large groups of targets in wide battlefields are critical inputs to formulating sound military decisions. Modern airborne radar sensors can provide wide-area surveillance coverage of battlefield ground activities. When obscured by terrain or other factors, some objects may only be detectable at intervals, generating intermittent radar data and creating difficulties for tracking groups over time. We present an algorithm, termed CoreTracking, which dynamically groups individual targets into clusters and tracks the clusters over time. Most traditional clustering techniques are static-object-oriented. We propose a "core member" concept to support dynamic-object-oriented clustering and to mitigate the effects of data intermittence. Observing the movement of the core cluster members, we can track the clusters across frames and predict their future movements. The performance and results of applying the CoreTracking algorithm to CASTFOREM data sets is also presented.