Model-based line-of-sight detection of an irregular celestial body for autonomous optical navigation

This paper proposes a model-based line-of-sight (LOS) determination system with high-precision for irregular celestial bodies applied in autonomous optical navigation. The 3D model of the target celestial body is assumed to be given in the task. The proposed system utilizes the 3D model to construct a 2D figure feature database. These 2D figures are extracted from regular sampled viewpoints in a sphere surface, and the selected features represent the contour shape of the figures and the corresponding LOS information. Then, a LOS determination system is developed based on the 2D figure feature database. The system detects the LOS unit vector of a given image though three steps. Firstly, the contour features of the target celestial body are extracted from the given image. Secondly, the best matching template is searched out from the database based on shape similarity of the contour. Thirdly, the LOS unit vector of the given image is generated by using the LOS information of the searched template based on similarity principle. Experiment results demonstrate the effectiveness and advantages of the proposed method.