Abstract: In order to meet the autonomous navigation requirements of asteroid approaching impact mission, a method of dim target asteroid recognition based on image registration was proposed. Due to the small size and weak brightness (usually above 10 magnitude) of asteroids, navigation sensors were required to have the ability to image dim targets. This caused the navigation sensor to simultaneously capture a large number of unknown dim stars, posing a challenge to the accurate identification of the target asteroid. The paper utilized the relative motion of asteroids and background stars. Firstly, a combination of ORB feature point localization and BEBLID feature point description was used to register inter-frame images. Secondly, star points were identified based on threshold segmentation, and the structural similarity index between corresponding windows was calculated for each star point. Finally, the detection of dim target asteroids was completed. Compared with the traditional image registration method and target asteroid detection methods, this method has improved speed and accuracy, has overcome the problems of dim target asteroid and unknown background stars, and provides such information as sight vector for optical autonomous navigation of asteroid defense.