Page 239 - Mechatronics for Safety, Security and Dependability in a New Era

Ch46-I044963.fm Page 223 Tuesday, August 1, 2006 3:57 PM
1, 2006
Tuesday, August
Page 223
Ch46-I044963.fm
3:57 PM
223
223
VISION-BASED NAVIGATION OF AN OUTDOOR
MOBILE ROBOT USING A ROUGH MAP
Jooseop Yun, Jun Miura and Yoshiaki Shirai

Department of Mechanical Engineering, Osaka University,
Suita, Osaka, 565-0871, Japan

ABSTRACT

We describe a method of mobile robot navigation based on a rough map using stereo vision, which
uses multiple visual features to detect and segment the buildings in the robot's field of view. The
rough map is a map with large uncertainties in the shapes and locations of objects so that it can be
built easily. The robot fuses odometry and vision information using an extended Kalman filter to
update the robot pose and the associated uncertainty based on the detection of buildings in the map.
An experimental result shows the potential feasibility of our localization method in an outdoor
environment.

KEYWORDS
Outdoor mobile robot, Vision-based navigation, Rough map.

INTRODUCTION
In this paper, we deal with the case that the robot has an environment map to be represented as a set of
2D segments. The map approximates the outlines of buildings except for feature information to be
used as landmarks (Georgiev, et al. 2002). We propose a method to robustly estimate the robot pose
using multiple visual features: walls of buildings, vanishing points, and corners of buildings. The
walls of buildings are extracted from the stereo vision observation. The vanishing points are calculated
from the non-vertical skylines of buildings. And the corners of buildings are the vertical skylines. The
visual features are matched to the given map and the results are integrated into the odometry
information for the estimation of the robot pose using an Extended Kalman Filter.

FEATURE DETECTION
For the matching process, we use multiple visual features: walls of buildings from disparity image,
vanishing points from non-vertical skylines, and corners of buildings from vertical skylines.

Walls of Buildings