Directional DIC method with automatic feature selection

Thijs Masmeijer

doi:10.1016/j.ymssp.2024.112080

2025-02-01

PRE

DOI：10.1016/j.ymssp.2024.112080

原文链接

原文求助

摘要

En 中文

Using displacement identification via high-speed cameras is a non-contact, full-field technique for characterizing the dynamic properties of engineered structures. Achieving reliable measurements of structural deformation through conventional Digital Image Correlation (DIC) requires a well-defined spatial gradient in two orthogonal directions within an image subset of pixels. However, DIC is susceptible to the aperture problem, which leads to an ill-conditioned optimization problem. An example of such a case is cable systems on which no speckle patterns can be applied. This study proposes Directional DIC (D-DIC) as an alternative technique to overcome the limitations posed by the aperture problem in conventional DIC methods. Directional DIC assumes the local direction of motion, which is often a valid assumption in structural vibration tests, as individual vibration shapes are often unidirectional at localized parts of the structure. This alleviates some restrictions on trackable image subsets, enabling more trackable locations. A parameter for quantifying expected tracking performance for DDIC is introduced. The parameter enables automatic feature selection for D-DIC, making optical methods more accessible for displacement identification on structures where no speckle patterns can be applied. Experimental modal tests are conducted using a flexible spider web-like structure to validate the D-DIC method and compare it to conventional DIC. The experimental findings show that the Directional DIC method allows for measuring displacements at significantly more locations than conventional DIC. Additionally, D-DIC provided less noisy frequency response functions, more retrieved stable poles, and more fitted vibrational modes. Lastly, the findings show that D-DIC is especially superior to DIC when using small subsets since D-DIC is less inhibited by the aperture problem.

Keyword:

Image-based structural identification

DIC

Directional DIC

Optical flow