Vision-based localization for leader-follower formation control