A-Track is a tracking process that is driven by automatic target recognition techniques (ATR), mobility analysis using digital maps and exploitation of other constraining information. It is structured as an optimization problem amenable to the formalism from classical relaxation labeling algorithm. The novel combination of techniques in A-Track provides a new approach to the temporal problem of establishing and extending target tracks through a sliding time window involving a sequence of multiple data frames. A data frame is a collection of sensor reports taken from one scan of a predetermined surveillance area that is being systematically and repeatedly scanned over time. The relative time differences between reports within a frame are generally much smaller than the time difference between different frames. A-Track is especially designed to handle cases having relatively large time intervals between frames that cause problems for conventional tracking algorithms based on predictive models.
CROSS REFERENCE TO RELATED APPLICATION
This application is a continuation-in-part of U.S. patent application Ser. No. 10/370,307, filed 10 Feb. 2003 now abandoned, which is hereby incorporated by reference.
