Autonomous spacecraft navigation using computer vision: a case study for the moon

Planetary space exploration by unmanned missions strongly relies on automatic navigation methods. Computer vision has been recognized as a key to the feasibility of robust navigation, landing site identification and hazard avoidance. We present a scenario that uses computer vision methods for the early identification of landing spots, emphasizing  the phase between ten kilometers from ground and the identification of the lander position relative to the selected landing site. The key element is a robust matching procedure between the elevation model (and imagery) acquired during orbit, and ground features observed during approach to the desired landing site. We describe how (1)
 preselection of characteristic landmarks reduces the computational efforts, and (2) a hierarchical data structure (pyramid) on graylevels and elevation models can be successfully combined to achieve a robust landing and navigation system. The behavior of such a system is demonstrated by simulation experiments carried out in a laboratory mock-up. This paper follows up previous work we have performed for the Mars mission scenario, and shows relevant changes that emerge in the Moon mission case of the European Space Agency.