A vision-based navigation system is considered to be essential for space activities occurring in near distance. In case of operations such as autonomous rendezvous and docking, the level of accuracy becomes much stricter than it is in separate orbit, so it is necessary to equip an additional navigation system for safety. In this study, we developed a vision-based navigation system which follows a straightforward algorithm of sequential image processing. A ground-based formation flying testbed is utilized to experimentally evaluate the performance of the navigation system. Based on the picture of beacon taken by the optical camera, the relative attitude between two simulators is estimated and compared with the results provided from the motion capture system to verify the level of precision.