Vision-based Robot Localization Across Seasons and in Remote Locations

Download Report

Transcript Vision-based Robot Localization Across Seasons and in Remote Locations

Vision-based Robot Localization Across
Seasons and in Remote Locations
Anirudh Viswanathan, Bernardo Pires, and Daniel Huber
The Robotics Institute, Carnegie Mellon University, USA
• GPS-denied unmanned ground
vehicle (UGV) localization by
matching ground images to a
satellite map
• Maps captured in different seasons
of the year exhibit high variation in
appearance caused by changes in
vegetation, etc.
• Image-matching using semantic
information, invariant to seasonal
change, is used to localize the UGV
• Localization across seasons is
demonstrated for maps captured in
spring, summer, and winter
Localizing ground-based panorama to
maps captured in different seasons