Abstract: Abstract： Focusing on the problems of computational burden and shortage of storage space for visual navigation in extraterrestrial body landing missions, a sparse descriptor for landmarks is proposed in this paper. The scale prediction and scale transformation of descent images are introduced, and the image features are extracted based on Harris algorithm, and FREAK algorithm is improved to descript the image features. Finally, the performances of the proposed algorithm, FREAK algorithm and SURF algorithm are compared under four external disturbances. The simulation results show that the proposed algorithm greatly reduces computational burden and demand for storage space, and is more suitable for landing missions of extraterrestrial objects.

Key words: Keywords： extraterrestrial landing, landmarks, sparse descriptor, improved FREAK