Intelligent transportation system technologies typically use inductive loop and magnetometer sensors to detect cars and trucks, but the sensors are not effective at detecting and classifying motorcycles and bicycles. Sensors detect the effects that electrically conductive materials have on electromagnetic fields, but both motorcycles and bicycles have low conductive masses. Loop detectors require that bicyclists be near the pavement markings for the sensors to trigger, and the detectors' sensitivities are difficult to set to avoid false detections. Accurate detection is critical for intersection control devices to ensure that traffic signals respond appropriately to motorcyclists and bicyclists. This paper reports on a multiphase research project to develop a system to detect and classify two- and three-wheeled vehicles more effectively. Phase I of the project, now complete, focused on developing an accurate detector. The result is a multi-instrument device for detecting these vehicles at intersections and ultimately for improving safety for riders. The need for an over-roadway sensor that works in a variety of weather, lighting, and time-of-day conditions led the researchers to select a multiple-technology rather than single-technology sensor. The researchers chose an infrared (IR)-visible light stereo camera to identify the riders on two- and three-wheeled vehicles, an IR thermal camera to distinguish cars and motorcycles from bicycles, and an acoustic sensor to distinguish classes of cycles such as large motorcycles versus mopeds. Phase I testing showed that the performance of the multi-sensor motorcycle classifier was promising, even though it mis-classified vehicles on several occasions. Development is underway to reduce the motion blur and improve the acoustic features that caused these misclassifications. The researchers will continue to refine the tool in the second phase of the project, now underway with a 24-month project period. (Author/publisher)
Abstract