A non-contact measurement technology based on binocular vision has been proposed for fault diagnosis. The inverse depth-based Bundle adjustment (BA) algorithm is used in nonlinearly optimizing the external parameters of the binocular vision system, thereby improving the measurement accuracy for large field-of-view applications. The study on damage detection in a turbine blade is also explored based on experimental deflection measurements. By fitting the rotation plane of the turbine blade, the normal deformation of the rotary blade is extracted through the coordinate transformation. And the location of the damage in the blade is then determined by comparing the difference in peak to peak magnitudes between blades with and without cracks. An experimental wind turbine model was built with blade length of 2.15 m. The 3D displacement distribution of the rotating blade was characterized, and the prefabricated cracks in the blade were also detected. The feasibility of the proposed methods is preliminarily verified with the satisfactory measurement results.