This study proposes a novel signal processing method integrating Digital Image Correlation (DIC) with wavelet packet analysis to achieve operational modal identification of structures under non-stationary excitation. The methodology systematically details the acquisition of structural vibration responses via DIC, enhancement of signal quality through wavelet packet-based noise reduction and stabilization, and subsequent extraction of modal parameters using the Polymax (stochastic subspace identification) method. A finite element model of a fixed-ended beam was developed for benchmark modal analysis, yielding theoretical reference parameters. Experimental modal analysis using conventional sensors validated these theoretical results, establishing the structure’s true modal characteristics. Subsequently, DIC experiments captured multi-channel displacement time histories. Application of the proposed method to this DIC data effectively identified modal parameters, even under non-stationary conditions. Experimental validation revealed close agreement: errors between the identified and reference natural frequencies were merely 3.3% for the first bending mode and 0.6% for the third bending mode. The proposed DIC-wavelet packet-Polymax approach demonstrates significant advantages in suppressing background noise and enhancing the accuracy of operational modal analysis for structures experiencing non-stationary loads.