Map-Enhanced Detection and Tracking from a Moving Platform with Local and Global Data Association

We present an approach to detect and track moving objects from a moving platform. Moreover, given a global map, such as a satellite image, our approach can locate and track the targets in geo-coordinates, namely longitude and latitude. The map information is used as a global constraint for compensating the camera motion, which is critical for motion detection on a moving platform. In addition, by projecting the targets' position to a global map, tracking is performed in coordinates with physical meaning and thus the motion model is more meaningful than tracking in image coordinate. In a real scenario, targets can leave the field of view or be occluded. Thus we address tracking as a data association problem at the local and global levels. At the local level, the moving image blobs, provided from the motion detection, are associated into tracklets by a MCMC (Markov Chain Monte Carlo) Data Association algorithm. Both motion and appearance likelihood are considered when local data association is performed. Then, at the global level, tracklets are linked by their appearance and spatio-temporal consistence on the global map. Experiments show that our method can deal with long term occlusion and segmented tracks even when targets leave the field of view.