Image Analysis-Based Study on Influencing Factors and Reliability in Soil Dynamic Deformation Time-History Identification

Abstract

Image deformation analysis techniques are widely used in geotechnical engineering. However, research on identifying the time history of dynamic soil deformation remains limited, and the reliance on empirical parameter selection often leads to unstable identification accuracy. To evaluate the applicability of these techniques for dynamic deformation time-history identification and to determine the key influencing factors, this study employs both the Reliability-guided digital image correlation and PIVlab methods on digitally simulated images of sandy soil. It systematically analyzes the effects of strain calculation window size, image acquisition frame rate, and deformation magnitude on the identification accuracy of displacement, acceleration, and strain time histories. The reliability of these techniques is also validated. The results indicate that under constant strain conditions, the calculation error decreases with an increase in the predefined strain value. Accordingly, the required minimum window sizes are no less than 17 × 17, 13 × 13, and 5 × 5 for strain levels of 0.1%, 1%, and 10%, respectively. Under sinusoidal loading, a frame rate equal to or exceeding six times the loading frequency can limit the displacement and acceleration errors to within 10% and 20%, respectively. For specific seismic waveforms—Northridge, Loma Gilroy, and Tangshan—the minimum frame rates required are 20 Hz, 10 Hz, and 6 Hz, respectively. When the deformation amplitude is ≥0.1 mm, the errors associated with both methods are comparable. However, for amplitudes ≤ 0.01 mm, PIVlab demonstrates superior accuracy in displacement identification. Notably, under large deformation conditions, PIVlab is prone to accumulating outliers, and its errors are significantly amplified upon differentiation to obtain acceleration. This study establishes key parameter thresholds and elucidates the technical characteristics of different methods, providing valuable references and guidance for soil dynamics experiments and the engineering measurement of dynamic deformations.