Author: Philip Morse

Publisher:

Published: 2019

Total Pages:

ISBN-13:

"As technology within the transportation industry develops, road users are presented with increasingly complex information. Measuring the impacts of driver feedback devices is a constant challenge for researchers and road network authorities alike. The objective of this thesis is to examine dynamic road user feedback systems and expand on automatic video-based data collection methods for evaluating these systems using new computer vision technology. This is accomplished through three different studies. First, a systematic literature review on variable message signs (VMS) is undertaken. Topics explored include a review of variable message sign study methodologies, legibility factors, operational factors and measures of effectiveness. Second, a dynamic speed display sign (DSDS), which uses many parameters measured in the first study, is evaluated in a before and after study on a low-speed arterial (30 km/h) with low traffic volumes with excessive speeding problems. Video-based data collection is used to obtain vehicle speeds. Improvements of 9.5 km/h (19.2%) for speed and 60% and 45% for fatality risks of vehicle-pedestrian and vehicle-cyclist collisions are measured. Thirdly, to improve from manual annotation to automatic, the transferability of video tracking optimization between camera views for traffic data collection is evaluated. Ideal parameters are obtained through use of a genetic algorithm with a MOTA fitness function. Optimized parameters were found to transfer well under clear condition, with performance results that ranged from 88% to 98% of their maximum. However, the winter, low-visibility video, needed an entirely different set of requiring much higher computation time." --