Broad Area Colloquium For AI-Geometry-Graphics-Robotics-Vision
(CS 528)

Computer Vision for Mars Exploration

Larry Matthies
Supervisor, Machine Vision Group
Jet Propulsion Laboratory
Monday, April 5, 2004, 4:15PM
TCSeq 200
http://graphics.stanford.edu/ba-colloquium/

Abstract

The JPL Machine Vision Group conducts research under funding from NASA, DARPA, and the U.S. Army on perception systems for autonomous navigation of unmanned ground and air vehicles (UGVs and UAVs). In this seminar, I will concentrate on NASA applications in planetary exploration, particularly for Mars. I will start by describing our contributions to the Mars Exploration Rover (MER) mission now operating on Mars. For this mission, we developed (1) the Descent Image Motion Estimation System (DIMES) for estimating horizontal velocity during terminal descent, which was used in retro-rocket firing logic, (2) onboard stereo vision and obstacle avoidance algorithms for rover navigation, and (3) onboard, stereo vision-based visual odometry algorithms to improve rover position estimation. I will then outline key directions and results to date in vision systems for future planetary exploration applications, including pin-point landing based on recognizing crater landmarks during descent and autonomous landing hazard avoidance using onboard structure-from-motion, which we have demonstrated on a robotic helicopter. Finally, I will show a few highlights from non-NASA work, particularly vision-based moving object detection on-the-move for ground robots.

About the Speaker

Larry Matthies received a PhD in computer science from Carnegie Mellon University in 1989 and has been at the Jet Propulsion Laboratory since then. He has led the Machine Vision Group there since it was formed in 1997. His research interests are in perception for autonomous navigation of unmanned ground and air vehicles, including 3-D perception, motion estimation, and terrain classification for day/night, all-terain, all-weather operation. He participated in development of the structured light range-finding system that was used in the 1997 Mars Pathfinder mission and he developed the stereo vision and visual odometry algorithms that are in use on Spirit and Opportunity in the 2004 Mars Exploration Rover mission. He is also Adjunct Professor in the Computer Science Department at the University of Southern California.

Contact: bac-coordinators@cs.stanford.edu

Back to the Colloquium Page