Abstract: To implement highprecision relative navigation during the descent and landing phase of asteroid exploration, a landmarkbased relative navigation method is proposed. The predicted image of the landmark map(Lmap/maplet) is first rendered by the stereophotoclinometry (SPC) illumination model, and the extracted image of the maplet is then generated from the observed image using bilinear interpolation. The combined correlation matching method is adopted to perform exact matching between the predicted image and the extracted image, and obtain the observed image location of the landmark center accurately. Therefore, the camera pose is estimated by the perspectivenpoint (PNP) algorithm. The simulation results show that this method can achieve subpixel accuracy even when the initial camera pose deviation is slightly large, and it has good realtime performance.

Key words: asteroid exploration, landmark navigation, image matching, subpixel accuracy