Aiming at the problems that the single vision positioning of the boat-mounted UAV is easy to be affected by the interference of marine environment, and it is very easy to have identification code loss, positioning error and large dead zone at the near-boat during autonomous landing, an unmanned boat-mounted UAV autonomous landing method with the cooperative fusion of sea and air vision positioning was proposed. By carrying cameras on the UAV and the UAV landing platform and using the cooperative identification codes laid on the landing area of the UAV and the bottom of the UAV, the co-operative navigation of sea-air dual vision was carried out, while the Kalman filtering algorithm was used to predict the target position, which not only improved the efficiency of identification but also predicted the position of the target in case of brief loss, and on the basis of the co-operative precise positioning of sea-air vision combined with PID, the unmanned boat-mounted UAV can be controlled by the PID control algorithm. The results of simulation experiments and at-sea experiments demonstrate the feasibility and effectiveness of the visual cooperative fusion localization algorithm and autonomous landing control method based on air-sea cooperation proposed in this paper, which can provide an effective solution for the autonomous recovery of heterogeneous class of submarine air-sea robotic systems.