We present a monocular vision-based road find-
ing navigation system that incorporates two contrasting ap-
proaches: region segmentation that estimates the road ap-
pearance, and road boundary detection using vanishing point
estimation. Our algorithm works in a general road setting,
and requires no training or camera calibration to enable
quick deployment. We tested our system in 1 indoor and 3
outdoor environments for a total 606.69m or 8375 frames. On
average, it is able to correctly estimate the center of the road
within 22.65cm. In addition, we also employ the system in
our ground-based robot, Beobot 2.0, for real-time autonomous
visual navigation. In 20 trial runs the robot was able to travel
autonomously for 98.19% of the route.
http://ilab.usc.edu/~kai/
Link to this comment:
All Comments (0)