Performance analysis of processing load distribution in camera networks for multi-target tracking

Camera networks have been widely used in surveillance and environmental monitoring in recent years. Among different topologies, there is a tendency to distributed topologies due to their efficiency in resource usage. But the comparatively low processing capability of individual smart cameras is an obstacle to run time-consuming algorithms like multi-target tracking on them. One suggestion is to share the extra processing load of busy nodes through the others, by taking into account the networking delays and different application requirements. In this work, we define and formulate the optimality, and analyze the optimum sharing amount of overloads for different conditions using queuing theory. Experimental results confirm that the analysis are valid and satisfy pre-defined requirements.