Assessment and monitoring of bridges using various camera placements and structural analysis

Abstract

This study presents a new vision-based deep learning method to monitor and evaluate the structural health of in-service infrastructure. For this purpose, three different camera placements, including remote, structure-mounted, and drone-mounted cameras, are proposed to capture the vibrations or displacements of bridges. The vision-based deep learning method is verified by an optical flow approach. Various techniques, such as visual data denoising and camera motion removal, are utilized to process the test data for displacement measurements and extract the structural frequencies. Structural models of bridges are analyzed to validate the measurements and assess the structural health of several pedestrian, traffic, and railway bridges without interfering with traffic. Measurements in the field experiments and results from the structural analysis on tested bridges show that the proposed framework works successfully and can be potentially engineered to monitor the structural health of existing bridges.