PSI - Issue 84

Francesco Allegrezza et al. / Procedia Structural Integrity 84 (2026) 81–88

85

Fig. 3. Setup employed and ROI.

To isolate and analyse the error intrinsic to each algorithm, post-processing was performed on the acquired signals. In particular, the tester signal was resampled and aligned to the measurements derived from the camera. A scale factor correction was introduced to account for uncertainties in the physical dimensions of the printed target and in the estimation of its side length in pixels. The error signals were then detrended by removing their mean value, to compensate for the constant offset associated with the initial position of the target within the image frame. The results, shown in Fig.4, highlight the presence of a systematic error also in the real case, in agreement with what was observed in the synthetic experiments. In the absence of blurring (Fig.4a), this error manifests a spatial periodicity related to the mm/pixel conversion factor, confirming the unit periodicity observed in the simulations. In addition to this contribution, it is possible to identify a lower frequency error component. When blur is introduced (Fig.4b) the systematic error is significantly reduced for both algorithms. For the algorithm based on quadratic interpolation the remaining error is dominated by the lower frequency component, whereas the upsampling approach exhibits a residual error with larger magnitude.

Fig. 4. Error comparison between algorithms.