With the development of image processing technology and the progress of embedded hardware, machine vision of unmanned aerial vehicle (UAV) has become a very hot research field. Automatic landing control of UAV is one of the key technologies of UAV flight control system, which plays an important role in the stability, accuracy, reliability and real-time performance of UAV landing. Aiming at the shortcomings of vehicle-mounted UAV with large error in automatic return and landing, through the airborne visual processing unit, the landing platform of the UAV is processed by using image processing algorithm, and the acquired location information of the UAV and landing platform are converted into real coordinate information byusing the coordinate conversion algorithm, and then the UAV is controlled to achieve precise landing by simulating the remote control lever. Experiment shows that the UAV landing platform can be accurately identified by using the proposed method. Compared with the experimental results of only GPS positioning, this method can effectively increase the accuracy of UAV landing and has certain practical application value.