Geospatial Exploitation of Motion Imagery

Motion imagery is emerging as a major source for intelligence information, especially in rapidly evolving tactical deployments. Novel deployment techniques, e.g. video sensors onboard unmanned aerial vehicles, distributed sensor networks, and handheld and fixed position monitoring systems are supporting the collection of timely, geospatially-registered information, enabling the precise monitoring of time dependent activity. The transition from static to motion imagery introduces significant challenges - indexing and retrieving large amounts of data, and extracting spatiotemporal activity patterns. Here, we present an innovative spatiotemporal helix approach to deriving temporal geospatial information from motion imagery. We present our approach for the scenario of identifying and labeling of object trajectories, and discuss its generalization to other geospatial information. In contrast to earlier work, our approach is applicable to arbitrary motion trajectories, and leverages metadata imbedded in the motion imagery stream for robust recognition. Our approach is fairly robust to incomplete and noisy data as our representation performs spatiotemporal smoothing, and filtering operations. We also propose to develop uncertainty management extensions to provide motion imagery time and position error budgets. Our Phase I research will show the enhancements and expansions to existing motion imagery capabilities in Army topographic terrain units, military intelligence units, and NIMA.