Abstract
Traffic congestion can be reduced with traffic signal control systems that determine signal parameters based on future traffic volume forecasts, which rely on data from upstream intersections. Image processing vehicle detectors have seen recent use due to cost effectiveness and maintainability, but occlusion occurs with conventional detectors, resulting in detectionfailures. In response, a vehicle tracking algorithm was developed based on the Spatio-Temporal Markov Random Field (S-T MRF) model. This algorithm enables one camera to measure traffic volume in all directions at intersections. The details of the image processing vehicle detector for intersections and field test results are described.