Summary: Tracking of moving objects in video through invariant features in
their graph representation
ArtiVision Technologies Pte Ltd
96 Robinson Road #13-02, Singapore
School of Computer Science, Tel Aviv University
Tel Aviv 69978, Israel
The paper suggests a novel contour-based algorithm for tracking moving objects in a video
sequence. The algorithm assumes that the moving objects in the input frames are segmented.
Each segmented frame is transformed into region adjacency graphs (RAG) data structures. Based
on the frame segmentation result, the object's contour is divided into subcurves while junctions of
the contour are derived. These junctions are considered as unique "signature" of the tracked object.
Junctions from two consecutive frames are matched. The motion of the junctions is estimated in
a search area that is represented by the RAG edges in the consecutive frame. Each pair of the
matched junctions may be connected by several paths (edges) that are considered as candidates