A Graph-Probabilistic Model of Inter-Camera Association of Vehicles in Distributed Road Video Surveillance Systems

This paper presents a graph-based probabilistic approach to multi-camera vehicle tracking in road surveillance networks. Cameras are modeled as graph vertices, while admissible inter-camera transitions are represented as directed edges between exit and entry zones. Each transition is described by base and temporal probabilities, enabling candidate filtering before appearance-based matching. Experimental results show that proposed model reduces search space of inter-camera association while preserving true candidates in most cases. Approach provides an interpretable foundation for topologyaware vehicle tracking in urban camera networks.