Overview
The Synopses Generator provides online, summarized representations of trajectories of vessels and aircraft for the maritime and aviation use case, respectively. We assume that online surveillance data is arriving in a streaming fashion concerning positions from a set of such moving objects monitored over a large geographical area. Instead of retaining every incoming position, we propose to drop any predictable positions (along trajectory segments of “normal” motion characteristics) with minimal loss in accuracy. Effectively, we may keep only those positions conveying salient mobility features (stop, changes in speed, heading or altitude, etc.) identified when the pattern of movement for a given object changes significantly. This derived stream of trajectory synopses must keep in pace with the incoming raw streaming data so as to get incrementally annotated with semantically important mobility features once they get detected. Such lightweight trajectory synopses will progressively become data-at-rest for permanent storage and may be exploited by other modules in further processing, such as contextual enrichment, recognition and forecasting of complex events, visual and mobility analytics, statistical analysis, mining.
We have consolidated a stream-based methodology for effective and efficient maintenance of trajectory synopses tailored for moving vessels and aircrafts. Moreover, we have specified parametrized characterizations of primitive mobility features that must be identified online, as well as suitable heuristics that can eliminate inherent noise and can provide succinct trajectory representations in one-pass over the incoming streaming positions. The final prototype implementation of the Synopses Generator has been tested against various real-world and synthetic datasets for both use cases. This comprehensive experimental study offers concrete evidence about performance, scalability, compression efficiency, and robustness of this framework and shows its strong potential for effective, real-time mobility surveillance. Moreover, the quality of the resulting trajectory approximations is very satisfactory.