Abstract
Abstract. Vision-based navigation has become an attractive solution for autonomous navigation for planetary exploration. This paper presents our work of designing and building an autonomous vision-based GPS-denied unmanned vehicle and developing an ARFM (Adaptive Robust Feature Matching) based VO (Visual Odometry) software for its autonomous navigation. The hardware system is mainly composed of binocular stereo camera, a pan-and tilt, a master machine, a tracked chassis. And the ARFM-based VO software system contains four modules: camera calibration, ARFM-based 3D reconstruction, position and attitude calculation, BA (Bundle Adjustment) modules. Two VO experiments were carried out using both outdoor images from open dataset and indoor images captured by our vehicle, the results demonstrate that our vision-based unmanned vehicle is able to achieve autonomous localization and has the potential for future planetary exploration.
Highlights
Accurate and autonomous localization of spacecraft and rover is essential for planetary exploration
The optical camera has the characteristics of low cost, wide application range, low power consumption and large field of view compared with the other detection devices, vision-based navigation has become an attractive solution for autonomous navigation
We can draw the conclusion that ARFM-based VO system contributes to eliminate the accumulative error of translate error, and the accuracy of this platform is 0.53%, which can meet the requirement of planetary exploration
Summary
Accurate and autonomous localization of spacecraft and rover is essential for planetary exploration. Traditional navigation method based on ground remote control system cannot achieve autonomous navigation in terms of communication latency and bandwidth limitation. While autonomous navigation can be achieved by GPS on Earth, there is no such positioning system on other planets. The optical camera has the characteristics of low cost, wide application range, low power consumption and large field of view compared with the other detection devices, vision-based navigation has become an attractive solution for autonomous navigation. The combination of unmanned vehicle platform and robust computer vision algorithms is a challenge task. A binocular camera based unmanned vehicle platform has been built and an ARFM (Adaptive Robust Feature Matching) based VO (Visual Odometry) software system has been developed to achieve GPS-denied autonomous navigation for planetary exploration
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have