Low cost video-based traffic congestion monitoring using phones as sensors

Traffic congestion is a serious problem in developing-world cities, compounded by the fact that it is unpredictable. Better real-time information about congestion levels can help with more effective use of infrastructure, for example allowing journey planning based on predicted congestion levels, or on selecting optimal routes based on current congestion levels at different places. In this paper we describe hardware and image processing methods for constructing a vision-based traffic congestion monitoring system tailored to the constraints of monitoring chaotic developing world traffic. By designing roadside monitoring units around camera phones, our prototype radically reduces costs compared to convential CCTV systems and hence makes it practical for deployment in this context. We show results from this system operating in Kampala.