Detection and Tracking of Objects in Airborne Video Imagery

We address the detection and tracking of moving objects in a video stream obtained from a moving airborne platform. The approach is based on the compensation of the image flow induced by the motion of observation platform and the detection and tracking of moving regions. The use of such an approach leads us to deal with stabilization inaccuracies, false alarms and non detection of moving objects and tracking difficulties due to partial occlusion or stop and go motion. Our approach use a hierarchical, feature based stabilization scheme and normal component of the residual flow for detecting moving objects. These objects are tracked using a dynamic template for each object, and extracts trajectories as the result of a graph searching algorithm. The proposed framework shows that an integration of well known tools and an efficient description of the moving objects can give very accurate detection and tracking of moving objects.