Aiming at the disadvantage that manual detection of large-scale wind turbine structures is time-consuming， labor-intensive， and dangerous， an identifying method for dynamic characteristics of large-scale wind turbine structures based on the optical flow method and UAV is proposed. The vibration law of UAV hovering in space was first studied， and the space displacement compensation method of the UAV lens based on background fixed point and self-scale factors was proposed. Then the UAV was used to capture the vibration videos of the wind turbine structure in the edgewise and flap-wise direction， and the optical flow method was used to visually track the monitoring points to obtain its displacement response， which was compared with the traditional monitoring methods in time history domain and frequency domain. Finally， the global modal shapes of wind turbine structure were identified by segmented monitoring method and verified by shaking table test. The results show that the displacement compensation method of background fixed point and self-scale factors can effectively solve the influence of spatial displacement drift caused by vibration during UAV hovering. Combined with the segmented monitoring method， the computer vision monitoring system equipped on UAV can identify the mode shapes of large-scale wind turbine structures well， which can realize long-distance， non-contact， and low-cost structural health monitoring， and provide a basis for the state evaluation of large-scale structures such as wind turbines.