ESTIMATING THE QUEUE LENGTH TO OPTIMIZE THE GREEN TIME FOR AN URBAN TRAFFIC CONTROL SYSTEM

Abstract: We describe a queue length estimation module that is part of an Urban Traffic Control System (UTCS). The main objective of this module is to estimate the traffic queue length at an intersection using image processing algorithms. This measurement feeds a traffic green time optimizer that chooses adequate green periods, based on the traffic load, and reduces the global time spent by drivers. We compare the results of three algorithms that estimate the queue length based on Gaussian Mixture Models and Level Lines. A hybrid method, which combines both approaches, is better at minimizing the system errors. In addition, this framework makes it possible to test different camera positions to improve results