Data-Driven Bridge Weigh-in-Motion

Abstract

Heavy vehicle detection in the road system is now an urgent issue from the perspectives of law enforcement and road health monitoring. Weigh-in-motion (WIM) is a technology which estimates vehicle weights without stopping the vehicles. Paved WIM (PWIM) is expensive and has limited installation locations. Bridge WIM (BWIM), which utilizes bridge components as weight scales, is quite inexpensive and easier to install. BWIM requires the dynamic characteristics of the bridge and traffic conditions for accurate weight estimation. In general, such characteristics are measured by several experimental runs using a vehicle with known axle weights. The weighing accuracy may be greatly degraded owing to the influence of the other traveling vehicles. In this paper, we propose a data-driven BWIM using a neural network. The model parameters are optimized automatically by video analysis and vehicle identification between WIMs. The model can estimate vehicle weights accurately considering various traffic conditions that may degrade the weighing accuracy.