Dynamic prioritization of multi-sensor feeds for resource limited surveillance systems

Processing resource intensive operations for surveillance systems and taking appropriate measures in real-time is a big challenge. In multimodal surveillance system with limited computing resource, it is hard to execute all tasks or processes to identify events and suggest actions to the decision makers. To overcome this problem, we propose a dynamic task prioritization scheme that schedules surveillance tasks or activities according to surveillance context subject to resource availability. A utility analyzer is introduced for this purpose that determines usefulness of sensor data. Based on data utility, the proposed scheme dynamically prioritizes data streams of all sensors so that it can minimize event detection time from the time of its occurrence. Due to such regulation, our method can be used efficiently as an admission controller for real-time task scheduling. The simulation results indicate that the prioritization scheme achieves better system utilization and enhances surveillance functionality.