Activity detection in SAR CCD

Synthetic aperture radar (SAR) coherent change detection (CCD) images reveal subtle changes on the ground such as the ground disturbance caused by vehicle tracks. The automatic detection of vehicle tracks is a challenging problem as CCD images have numerous problems. In this paper, we focus on detecting likely activity, with the assumption that an activity of interest has one or more tracks. Even if the automatic track detector has many false alarms and missed detections, enough track segments are located to accurately detect activity. The novel contribution of this work is a mathematical framework that detects activity based on the spatial proximity of several individual track segments. Experimental results show a large improvement in the detection performance of images containing activity when the new method is employed.