AI-Assisted Infrastructure Monitoring: Supplementing Human Inspections on Auckland Harbour Bridge

Ensuring traffic safety on critical infrastructure such as the Auckland Harbour Bridge (AHB) is essential, particularly given its movable concrete barrier (MCB) system, which handles 154,000 vehicles daily on the AHB and helps manage traffic in similar scenarios in over 20 cities worldwide [1, 2]. The MCB system relies on connecting metal pins to secure 750 kg concrete segments together, and these pins can become dislodged due to external factors, posing significant safety risks. On-foot manual inspections are conducted to check for dislodged or at-risk pins, exposing workers to hazardous traffic and weather conditions. This research introduces a Spatio-Temporal Enhanced Network (STENet) framework [3] to automate the detection of unsafe pin positions. The system processes high-speed video data captured under diverse conditions, achieving 95.2% accuracy after incorporating synthetic data generated by cloning original video frames and manipulating their backgrounds.