This article presents the problems in traffic detection found in inductive loop detectors and monocular video detectors, and describes a newly-developed system, the Silhouette Vision Sensor (SVS) which uses a stereo-based video detector. This is seen as solving the previous trade-off problem between performance and easy set-up. The detector recognises a vehicle using its three-dimensional (3-D) information, and its performance is little affected by changes of weather and environmental conditions. SVS consists of a pair of 2-D images using camera intrinsic and extrinsic (road surface) parameters. Software makes the setting of parameters an easy task for even an untrained field engineer. Field tests have been carried out in Japan and the USA. For the covering abstract see ITRD E118381.
Abstract