Development of a segment-based description of events in imagesequences

摘要

This paper describes an approach for image sequence processing,which extracts objects (or parts of objects) from the image sequence,generates motion-specifically attributed tracks (location/time course)and derives a hierarchy of motion. The base consists ofobject-characterizing segments which are determined by a multiplethreshold method. This method yields feature descriptions for eachsegment. Based on this description, an association of the segments fromframe to frame is done. The resulting motion vectors are concatenated aslong as the features of the segments and the motion-specific attributesdo not change significantly. In a subsequent step the analysis ofmotion-specific interrelations between the tracks is executed. The eventrecognition is processed and controlled by event models. In conformitywith motion concepts the hypotheses of events are initialized by motiondependent changes of the tracks