Aiming at the problem of on-line measurement of vibration modes of large wind turbine blades, photogrammetry and covariance random subspace method are used to analyze the modes of large wind turbine blades at work. Firstly, the trajectories of key measurement points on the blades are obtained by photogrammetry. Then, the obtained data are analyzed by covariance driven random subspace, and the modal parameters of the tested blade are obtained, including natural frequency, natural damping and modal mode. Because field measurement is difficult, a 2MW wind turbine and wind field model were set up by GH Bladed, and the moving path of target points on the blade was simulated, and the modal parameters of the blade were calculated by using the proposed method. The experimental results show that the natural frequency relative error of the first to fifth order modes is less than 1.4%, and the relative error of the modes is less than 10.0%, which verifies the feasibility of photogrammetry and covariance driven random subspace method in the modal analysis of large wind turbine blades. The method is applied to a 3.5m blade wind turbine to obtain its natural frequency, which verifies the feasibility of photogrammetry and covariance driven random subspace method.