A STUDY ON AUTOMATIC MEASUREMENT OF PRECISE TRAFFICE ENGINEERING INDICATORS FOR TRAFFIC VOLUME BY INTERSECTION DIRECTION USING DEEP LEARNING

Abstract

Automatic traffic volume measurement systems using AI-based image recognition technology have been recently studied in the field of traffic engineering. Most of the systems for intersection use videos taken from high positions, but these systems are difficult to apply to intersections where it is difficult to install cameras at high positions. A method for the automatic directional traffic volume measurement at intersections using videos taken fromrelatively low locations (three meters above ground level) is proposed and achieved an F1 score of 0.97 to 0.99 for the directional traffic volume measurement. Furthermore, this paper pro- poses a method for automatic measurement of traffic volume by direction for each traffic light signal, which is difficult to be realized by conventional manual measurement, by simultaneously detecting the traffic signal light from the same video. The use of AI will not only replace conventional methods but also create new values in the field of traffic engineering.