The unstable vibration of the surface stress sensitive film sensing system in the wind tunnel brings large errors to the measurement results. A self-calibration method is proposed to perform vibration correction on the sensing images to eliminate the effect of unstable platform vibration and reduce the measurement errors. The initial displacement containing vibration and effective information is firstly obtained by cross-correlating the original sensing images, then the reference zone mask is obtained by Otsu thresholding method and morphology, and finally the whole vibration plane is restored and vibration corrected using the principle of perspective transformation and the least-squares method. The method involves the sub-zones of the sensitive membrane sensing image as feature points in the calculation, realizes the self-calibration correction of vibration, and enables the sensing system to perform high-precision measurement even on unstable platforms. The validity and practicability of the method are verified by high-speed train experiments, which show that the method has higher accuracy and better robustness than other traditional methods, produces good spatio-temporal resolution images, and reduces the vibration average error by 98% in an incoming flow of 50m/s.