Urban driving has become the focus of autonomous robotics in recent years. Many groups seek to benefit from the research in this field including the military, who hopes to deploy autonomous rescue forces to battle-torn cities, and consumers, who will benefit from the safety and convenience resulting from new technologies finding purpose in consumer automobiles. One key aspect of autonomous urban driving is localization, or the ability of the robot to determine its position on a road network. Any information that can be obtained for the surrounding area including stop signs, road lines, and intersecting roads can aid this localization. The work here attempts to combine some previously established computer vision methods to identify roads and develop a new method that can identify both the road and any possible intersecting roads present in front of a vehicle using a single color camera. Computer vision systems rely on a few basic methods to understand and identify what they are looking at. Two valuable methods are the detection of edges that are present in the image and analysis of the colors that compose the image. The method described here attempts to utilize edge information to find road lines and color information to find the road area and any similarly colored intersecting roads. This work demonstrates that combining edge detection and color analysis methods utilizes their strengths and accommodates for their weaknesses and allows for a method that can successfully detect road lanes and intersecting roads at speeds fast enough for use with autonomous urban driving.
Library of Congress Subject Headings
Computer vision; Color vision; Automated guided vehicle systems; Mobile robots
Department, Program, or Center
Computer Engineering (KGCOE)
Kurdziel, Michael Scott, "A monocular color vision system for road intersection detection" (2008). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus