Real Time ARM-based Traffic Level of Service Classification System

Traffic Level of Service (LOS) information is crucial for traffic management systems, especially in urban areas. One method to estimate traffic LoS is to use a central server system to process traffic images captured by road side cameras. However, this approach requires a high performance server system as well as high network throughput to transmit images from the cameras to the server, which results in very high system deployment cost. In this paper, we propose a cost effective distributed solution using smart cameras each of which is equipped with a low cost ARM microprocessor to estimate the LOS from the captured traffic images. The LOS of a road estimated by a corresponding camera will then be sent to a traffic information server. In this study, LOS is determined based on the average traffic flow speed and the road occupancy. The Lucas Kanade optical flow method is used to estimate the speed of the traffic flow. In order to have a real time processing on a low cost platform, the whole LOS estimation algorithm has been optimized. The experimental results show that our optimized implementation can process traffic images in real time on an ARM Cortex-A8 platform and is 4 times faster than an OpenCV based implementation on the same platform.