Non-contact vision-based deformation monitoring on bridge structures

Abstract

Information on deformation is an important metric for bridge condition and performance assessment, e.g. identifying abnormal events, calibrating bridge models and estimating load carrying capacities, etc. However, accurate measurement of bridge deformation, especially for long-span bridges remains as a challenging task. The major aim of this research is to develop practical and cost-effective techniques for accurate deformation monitoring on bridge structures. Vision-based systems are taken as the study focus due to a few reasons: low cost, easy installation, desired sample rates, remote and distributed sensing, etc. This research proposes an custom-developed vision-based system for bridge deformation monitoring. The system supports either consumer-grade or professional cameras and incorporates four advanced video tracking methods to adapt to different test situations. The sensing accuracy is firstly quantified in laboratory conditions. The working performance in field testing is evaluated on one short-span and one long-span bridge examples considering several influential factors i.e. long-range sensing, low-contrast target patterns, pattern changes and lighting changes. Through case studies, some suggestions about tracking method selection are summarised for field testing. Possible limitations of vision-based systems are illustrated as well. To overcome observed limitations of vision-based systems, this research further proposes a mixed system combining cameras with accelerometers for accurate deformation measurement. To integrate displacement with acceleration data autonomously, a novel data fusion method based on Kalman filter and maximum likelihood estimation is proposed. Through field test validation, the method is effective for improving displacement accuracy and widening frequency bandwidth. The mixed system based on data fusion is implemented on field testing of a railway bridge considering undesired test conditions (e.g. low-contrast target patterns and camera shake). Analysis results indicate that the system offers higher accuracy than using a camera alone and is viable for bridge influence line estimation. With considerable accuracy and resolution in time and frequency domains, the potential of vision-based measurement for vibration monitoring is investigated. The proposed vision-based system is applied on a cable-stayed footbridge for deck deformation and cable vibration measurement under pedestrian loading. Analysis results indicate that the measured data enables accurate estimation of modal frequencies and could be used to investigate variations of modal frequencies under varying pedestrian loads. The vision-based system in this application is used for multi-point vibration measurement and provides results comparable to those obtained using an array of accelerometers.