Vision-Based Simultaneous Localization and Mapping on Lunar Rover

Abstract

With the development of lunar exploration technology, vision-based localization and navigation technology has become a research focus in the field of lunar rover. This paper proposes an image-based method for localization and mapping with a lunar rover. The motion of the camera represents the movement of the lunar rover. Based on the images acquired by the camera, the relative pose of the camera and 3D landmarks are obtained using the multi-view geometry and the bundle adjustment optimization methods. The prior knowledge of the lunar rover movement is not required. In addition, this paper also proposes a grid-based feature extraction method to solve the problem of uneven feature extraction and mis-matching. The algorithm in this paper has been tested in real time in a large image dataset. Finally, the error analysis of the estimated pose obtained from the experiment and the real trajectory proves the excellent performance of the algorithm.