A comparative study of 3D reconstruction methods in stereo digital image correlation

摘要

The measurement accuracy of stereo digital image correlation (DIC) depends greatly on the accuracy of the reconstructed 3D point. There are four 3D reconstruction methods used in stereo DIC, such as least square method (LSM), optimal method (OM), mid-point method (MM), and geometrical optimal method (GOM). The 3D point from the OM is reconstructed by minimizing the re-projection error, while in the GOM the 3D point with the minimal re-projection error on the shortest line segment between the lines of sight is regarded as the desired one. However, no research has been conducted to compare the performances of these four 3D reconstruction methods in stereo DIC. In addition, it is generally believed that the accuracy of the 3D point having the minimal re-projection error is the highest, whereas, it is not validated in stereo DIC. Therefore, the objective of this study is to compare the performances of the four 3D reconstruction methods in stereo DIC in terms of measurement accuracy and computation efficiency. Simulation and experiment results demonstrate that the accuracies of the four methods are nearly the same and the lesser re-projection error may not imply the higher accuracy of the reconstructed 3D point if the matching accuracy of the corresponding points (CP) has already been determined. Finally, the LSM is recommended to be used in stereo DIC for 3D reconstruction due to its highest computation efficiency without compromising the measurement accuracy.