Work

Line segment tracking for camera-based vibration measurement of large structures in outdoor conditions

Year: 2025

Type: article

Abstract: This paper introduces and validates a computer vision method called the Line Segment Tracking (LST). It is designed for vibration measurement of large structures under outdoor conditions. Compared to conventional window-based methods such as Digital Image Correlation (DIC) and Lucas-Kanade Optical Flow (LKOF), the LST method minimizes disturbances within the region of interest by estimating the displacement of an intersection point determined by tracked line segments. First, error quantification and a parametric study of the LST method were conducted using synthetic images. The results were further validated with real images captured under laboratory conditions, utilizing a real-time camera orientation estimation program for precise alignment. Finally, the displacement of a tower crane in operation within an outdoor environment was measured. The validation of the LST method was performed by comparing its results with acceleration measurements. Comparative analysis demonstrated that the LST method not only produced results consistent with those obtained using DIC and LKOF but also effectively reduced the impact of light variations in images. In conclusion, the proposed LST method is a robust alternative to traditional window-based tracking techniques, suitable for vibration measurement in challenging outdoor environments. • Description of the Line Segment Tracking (LST) method for vibration measurement. • Validation of LST on synthetic and real images in lab and outdoor conditions. • Error and parametric analysis of LST with a real-time camera pose estimation program. • Measurement of the 3 dimensional displacement of a tower crane using LST. • Reduction of local light disturbance within ROI compared to window-based methods.

Source: Mechanical Systems and Signal Processing

DOI: 10.1016/j.ymssp.2025.113728

Authors: Yunhyeok Han, Gwendal Cumunel, Stefania Lo Feudo, Franck Renaud