Long-term monitoring system for full-bridge traffic load distribution on long-span bridges

Abstract

Long-term monitoring of traffic loads across the whole bridge is of great significance for bridge health monitoring and safety assessment, especially for long-span and complex bridges. However, none of the existing schemes have operated long term, and the full-bridge traffic load (FBTL) monitoring has remained a concept. In this paper, an improved FBTL monitoring framework capable of long-term operation was proposed, through the fusion of the weigh-in-motion system (WIMs) and multi-camera vision systems. In the system, vehicle detection and tracking algorithms with better accuracy and robustness were achieved with deep convolutional neural networks, which could effectively deal with the problem of target loss during long-term monitoring. A fast but accurate correction method for transverse vehicle position was proposed by applying projective geometry, ensuring the system could operate with a large number of target vehicles. The proposed FBTL monitoring system was successful in generating the traffic load distribution of a long-span cable-stayed bridge, demonstrating its technical feasibility and engineering applicability.