Unsupervised tracking algorithm for precise traffic estimation in panoramic scenes

Wu, Fangyu

Permalink

https://hdl.handle.net/2142/100926 Copy

Description

Title

Unsupervised tracking algorithm for precise traffic estimation in panoramic scenes

Author(s)

Wu, Fangyu

Issue Date

2018-04-05

Director of Research (if dissertation) or Advisor (if thesis)

Work, Daniel B.

Department of Study

Civil & Environmental Eng

Discipline

Civil Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Computer Vision
• Vehicle Tracking

Abstract

The traffic experiment conducted by physicist Sugiyama in 2007 has been a seminal work in transportation research. In the experiment, a group of vehicles are instructed to drive on a circular track starting with uniform initial spacing. The isolated experimental environment provides a safe, economic, and controlled environment to study free flow traffic and stop-and-go waves. This dissertation introduces a novel method that automates the data collection process in such an environment. Specifically, the vehicle trajectories are measured using a 360-degree camera, and the fuel rates are recorded via on-board diagnostics (OBD) scanners. The video data from the 360-degree camera is then processed by an offline unsupervised computer vision algorithm. To validate the data collection method, the technique is then evaluated on a series of eight experiments. Validation analysis shows that the collected data are highly accurate, with a mean position bias of less than 0.002 m and a small standard deviation of 0.11 m. The positional data also yields highly reliable velocity estimates: the derived velocities are biased by only 0.02 m/s with a small standard deviation of 0.09 m/s. Beyond the experimental methodology, the produced trajectory and fuel rate data can be readily used to study human driving behaviors, to calibrate microsimulation models, to develop fuel consumption models, and to investigate engine emission. To facilitate future research, the source code and the data are made publicly available online.

Graduation Semester

2018-05

Type of Resource

text

Permalink

http://hdl.handle.net/2142/100926

Copyright and License Information

Copyright 2018 Fangyu Wu

Owning Collections