Robust Map-Based Visual Localization System for Container Ports

PhD Thesis Proposal Defence


Title: "Robust Map-Based Visual Localization System for Container Ports"

by

Mr. Bohuan XUE


Abstract:

This proposal addresses the state estimation problem in Unmanned Ground 
Vehicles (UGV) systems at container ports by introducing a ground-feature-based 
vision positioning system tailored for dynamic environmental changes and 
interference from large metal structures. Innovatively, the system calibrates 
external parameters from the vehicle control center to the camera and 
dynamically adjusts these parameters in real-time in response to variations in 
container weight, ensuring high precision in positioning. The core of the 
system employs a deep learning and geometric approach to extract lane lines and 
ground diamond patterns, coupled with a robust outlier exclusion mechanism to 
ensure accuracy in state estimation. Moreover, a layout and automated 
construction scheme for map information is proposed to reduce manual 
intervention.

Compared to existing technologies, our system demonstrates significant 
improvements in error filtering, positioning accuracy, and computational speed, 
with a computation time reduction of 55ms and a 20.6% enhancement in 
translation precision. Deployed at the Nansha Port, the system has operated 
stably for over 20 months, illustrating its exceptional adaptability to the 
environment with an exceedingly low fault rate due to visual positioning.

Future work will focus on enhancing odometry accuracy, developing integrated 
position correction techniques, and exploring the synergistic use of vision and 
LiDAR technologies for augmented navigation. These advancements are anticipated 
to further improve the system's overall performance and capability to cope 
with complex scenarios.


Date:                   Friday, 15 March 2024

Time:                   4:00pm - 6:00pm

Venue:                  Room 5506
                        Lifts 25/26

Committee Members:      Dr. Yangqiu Song (Supervisor)
                        Dr. Dan Xu (Chairperson)
                        Dr. Xiaojuan Ma
                        Dr. Long Chen