Selecting ghosts and queues from a car trackers output using a spatio-temporal query language

This paper presents a spatio-temporal query language useful for video interpretation and event recognition. The language is suited to describe configurations of objects moving on a plane. To demonstrate its applicability it has been tested on the output of a tracker working on a car traffic scene. The results of two example sets of queries are shown in two videos generated from the trackers data output. The first selects a ghost from the tracking data and the second shows how to find queues of cars in the road traffic scene without prior knowledge of lanes.