A reliable transport infrastructure is vital to the commercial and lifestyle demands of a developed country, with the majority of journeys occurring by road. Bridges are a key component of this infrastructure, if a bridge fails or is unnecessarily closed it has widespread adverse eﬀects throughout the surrounding area. Detailed monitoring is essential to ensure adequate maintenance of these structures is carried out, this is not currently the case as many bridges are only sporadically checked by visual inspection, often by a junior engineer. Structural Health Monitoring (SHM) has been developed to counteract this shortfall, to date the instrumentation used has primarily been contact based and usually requires bridge closure. Computer Vision is the process of using cameras to obtain data from images, this method is now being applied to monitor civil structures worldwide. With regards to the monitoring of bridge displacement from applied vehicle load, the primary focus of existing research has been on single camera studies to monitor one point on the bridge without a means of identifying the cause of the measured displacement. The work presented in this thesis details the development of an accurate, time synchronised multiple camera solution for the monitoring of bridge displacement. The system has been validated for accuracy in numerous laboratory and ﬁeld trials against a diverse array of instrumentation and under a variety of environmental conditions. To facilitate load identiﬁcation from vehicles, a Deep learning based method for Vehicle Identiﬁcation has also been developed in the course of the work presented. The load identiﬁcation solution is capable of precise location and ﬁne-grained classiﬁcation of vehicles from images captured in millisecond level synchronisation with captured displacement readings. This composite system has been successfully veriﬁed in a ﬁeld trial, and with further development including incorporation of a weights database for approximate load calculation can provide the basis of a total system for bridge displacement monitoring.
|Date of Award||Jul 2020|
- Queen's University Belfast
|Sponsors||Invest Northern Ireland, Science Foundation Ireland & National Science Foundation|
|Supervisor||Su Taylor (Supervisor), Jesus Martinez-del-Rincon (Supervisor) & David Hester (Supervisor)|
- Computer Vision
- Deep Learning
- Structural Health Monitoring