Abstract :
Accurate traffic surveillance is vital to modern traffic control systems. Present vehicle detectors in widespread use are point detectors. That is, they detect a vehicle at a single point in space. Thus, a large array of point detectors is required to accurately monitor traffic in a given area. Recognizing the potential of an area type detector, the Federal Highway Administration (FHWA) contracted with the Jet Propulsion Laboratory (JPL) to investigate the feasibility of obtaining traffic flow parameter measurements automatically in real time from video images. JPL\´s experience in image processing has provided the expertise to design a "breadboard" version of a Sensor for Control of Arterials and Networks (SCAN) and demonstrate the feasibility of this concept. Synthesis of the vehicle detection and tracking algorithms and software has been the most challenging aspect of this effort. Wide ranging operational conditions such as shadows, reflections from wet pavement, varying sun direction, weather and night conditions have required a significant advancement in the state-of-the-art.
