Abstract: Aiming at the difficulty of relative positioning between the aerial robot and the grabbed target when the aerial manipulator system performs the tasks of grabbing, dropping, etc., a design method for vision-guided aerial manipulators is presented. Firstly, the overall mechanism design of the aerial manipulator system is introduced, and kinematics and dynamics models are developed for the aircraft and the aerial operation device. Secondly, the on-board monocular camera is utilized to detect the ArUco markers on the grabbed target according to the pinhole imaging model on the premise that the marker size is known, and the perspective-n-point (PnP) algorithm is used to calculate the camera pose. Then the flying robot and the operation device are controlled hierarchically based on the camera pose information. Finally, the effectiveness of the pose estimation algorithm is validated by the stationary and outdoor hover experiments, and the effectiveness and rationality of visual guidance are further verified by autonomously grabbing a tubular object with a diameter of 2 cm and mass of 100 g.