Freeway Travel Time Estimation using Existing Fixed Traffic Sensors ( Phase 2 ) – A Computer - Vision - Based Vehicle Matching Approach

The contents of this report reflect the views of the authors, who are responsible for the facts and the accuracy of the information presented herein. Travel time, one of the most important freeway performance metrics, can be easily estimated using the data collected from fixed traffic sensors, avoiding the need to install additional travel time data collectors. This project is aimed at fully utilizing the data gathered by existing fixed sensors to estimate point-to-point freeway network travel time in the MoDOT St. Louis District. The objectives of the project were achieved by accomplishing three major tasks: 1) A new travel time estimation model was developed and its ability to outperformed two traditional models, the instantaneous and time-slice models, demonstrated using real-world data. 2) The verification of the proposed model relies on the ground truth travel time. Collecting ground truth travel time is both time-consuming and challenging, so a novel Vehicle Re-identification (VRI) method was developed to facilitate the ground truth data collection process with satisfactory results. 3) A point-to-point network travel time estimation prototype system was also developed. In addition to the primary module of travel time estimation, both data assurance report production and traffic volume report production were modularized. The prototype system was then applied to four case studies measuring: freeway corridor performance (both with and without a turning junction), the impact of severe weather events on traffic volume, and travel time reliability. The prototype system clearly demonstrated its capability and efficiency through these case studies. Because of its high design flexibility, the system is confidently expected to support additional case studies with minimal system revision and tune-up.