A Micro-Vibration Measurement Method Combining Super-Resolution Network with Phase-Based Motion Estimation

Visual vibration measurement holds significant application value in industrial inspection and structural health monitoring, where phase-based methods have gained widespread attention due to their advantages in achieving robust sub-pixel displacement detection. However, images acquired in practical applications often suffer from motion blur issues, which limit the measurement accuracy of phase-based methods. To address this challenge, this paper proposes a phase-based vibration measurement method integrated with an improved Real-Esrgan super-resolution network to enhance the accuracy of vibration signal extraction. This method introduces a lightweight deblurring module (LDM) based on Real-Esrgan and constructs a high-order degradation model incorporating motion blur, enhancing the network's capability to restore typical blur phenomena in vibration videos. Experimental results demonstrate that after super-resolution preprocessing, the root mean square error of phase-based vibration measurement is significantly reduced, image edge details are effectively enhanced, and the system exhibits higher precision and robustness under various exposure conditions. This study provides an effective solution for high-precision non-contact vibration measurement in complex imaging environments.