Landmark selection and matching for aiding lunar surface navigation

The visual navigation on the lunar surface is a natural way to alleviate the accumulated error raised by the inertial navigation device. However, it is difficult to find distinctive landmarks for effective visual navigation due to the similar and textureless terrains on the lunar surface. In this paper, a novel algorithm is proposed to solve such a problem. Firstly, the local invariant features extract. Next, an improved density cluster technology employed is to initial landmark patches. Finally, we choose the most distinctive and robust patch by optimizing an objective function. For meeting the real time requirement, we also design an effective strategy to reduce the number of local invariant features. The proposed method evaluates in the lunar surface map with 50 meters accuracy. The experimental results exhibit satisfactory performance in terms of matching accuracy.