Test protocols for non-loop detectors have often required comparing the performance attributes of these detectors with those of loops or other point detectors, or to manual counts. However, that comparison is not always appropriate for a variety of reasons, and it does not provide all of the critical information needed to make acceptance or rejection decisions. In the case of video image vehicle detection systems (VIVDS), comparison with loops (i.e., simple count comparisons) provides only a limited glimpse on performance since the two systems have different perspectives on approaching vehicles. In all cases except those in which cameras are oriented vertically downward, cameras and loops or other pavement-based detectors detect vehicles at different points. Also, for VIVDS, factors such as the “aspect ratio” (ratio of horizontal distance to detection zones divided by the camera height) vary significantly, and these variables significantly impact the accuracy of camera-processor systems. The purpose of this document is to report on the development of a proposed concept for a VIVDS test protocol. (Author/publisher)
Abstract