Geospatial Exploitation of Motion Imagery

In Phase I, we studied a motion pattern modeling, identification and indexing method based on our innovative model of spatiotemporal helixes and demonstrated feasibility for application to urban scenes using video imagery. Our research to date has shown the feasibility of modeling video motion trajectories with spatiotemporal helixes, and using this model as an indexing mechanism for storage and query of events about motion trajectories. Our feasibility demonstration shows great promise in the efficiency afforded by our approach. Our model captures the point-mass motion trajectory with an optimal number of nodes. Similarly, our threshold-based approach captures motion dynamics with an optimal set of prongs. The similarity metrics that we developed will enable us to compare different trajectories - at the coarse level - with mobility state transition metrics, and - at a granular level - with spatiotemporal distance metrics. In this Phase II effort, we propose to develop the full-scale prototype implementation of a spatiotemporal helix based system for geospatial exploitation of motion imagery based on our Phase I research, evaluate its performance with laboratory and real-world data, refine the fundamental algorithms based on the evaluation, and transition the research to Phase III commercialization.