A Cost-Effective Template-Matching Vision System for Non-Contact Displacement Monitoring: Laboratory Validation Against Linear Variable Displacement Transducers

Abstract

Structural health monitoring increasingly relies on non-contact optical techniques to overcome the limitations of contact sensors such as linear variable displacement transducers (LVDTs) and accelerometers. This paper presents the development and assessment of a vision-based displacement measurement platform, built in Python with OpenCV, implementing the Template Matching Method (TMM) for automatic tracking of user-defined regions of interest (ROIs) in video sequences. The platform was validated in a controlled laboratory environment through two configurations: a small circular target of 81 mm diameter and a much larger ROI on the flanges of an HEB 300 steel section, both instrumented with a reference LVDT. Video was acquired with a digital camera positioned frontally at 240 cm from the monitored elements, at 59.94 fps. Displacement time histories from template matching were compared against synchronous LVDT recordings using the coefficient of determination (R2), mean absolute error (MAE), root mean square error (RMSE) and mean percentage residual. The small ROI yielded R2 = 0.9975, MAE = 0.89 mm and RMSE = 1.06 mm, while the larger ROI yielded R2 = 0.9932, MAE = 1.22 mm and RMSE = 1.51 mm. Both approaches required decontamination of the rigid-body motion of the supporting plate to which the LVDT was attached. Near-millimetre accuracy was obtained on both a marker-dominated and a texture-dominated ROI, at a fraction of the cost of alternative approaches reported in the literature. Beyond the present validation, the platform is offered as a self-contained, reusable research tool for displacement monitoring in other experimental configurations.