Machine‐Vision System for Multispot Vehicle Detection

Abstract

New technology and equipment are constantly sought by traffic engineers for dealing with the problem of urban congestion. Among the most promising concepts available today is the use of video imaging for vehicle detection, automatic surveillance, and extraction of data needed for developing advanced control concepts. In this paper, a recently developed video detection system is presented. The system operates in real time, it can detect traffic simultaneously at multiple points within the field of the camera's view, and it emulates loop detectors, i.e. it produces signals compatible with loops. The system was installed and tested both off‐line and in real time using taped data and field installations, respectively. The results up to this point suggest accuracy levels comparable to loops; higher accuracies than loops in speed measurements are expected. Finally, software is being developed for real‐time extraction of traffic parameters, state variables (i.e., queue lengths and size), and measures of effectiveness (delays, stops, energy consumption, etc.) by the device.