Abstract: Rotating machinery is widely used in critical fields such as aerospace, nuclear engineering, and power energy, where its vibration characteristics directly impact its service safety and stability. In the nuclear engineering sector, vibration measurement for specialized rotating machinery typically employs eddy current sensor technology, which has limitations on aspects of conductivity dependency, non-continuous displacement measurement when applied to new materials such as carbon fiber composites. In this study, a non-contact vibration testing technique for specialized machinery based on orthogonal sampling Moiré, combining binocular stereo vision was proposed to achieve full-field, high-precision, and measurement of three-dimensional displacement fields of complex curved structures under vibrations loading conditions. The method involves fully automated processing steps including epipolar correction, automatic marker point localization and tracking, phase analysis, phase matching, and 3D reconstruction, enabling the reconstruction of object surface 3D morphology and displacement fields to extract vibration time-domain and frequency-domain features. Experiments conducted on a 3D-printed cylindrical model under 5 Hz sinusoidal excitation demonstrated a reconstruction diameter error of 0.23% and a displacement field root mean square error (RMSE) reaching submicron levels. When being compared to traditional digital image correlation (DIC), the proposed method can significantly improves processing efficiency while maintaining accuracy.