Fatigue cracks developed under repetitive traffic loads are a major threat to maintaining the structural integrity of steel bridges. Human visual inspection is currently the de facto approach for fatigue crack detection. However, due to human limitations and the complex nature of bridge structures, fatigue crack inspections are time consuming, labor intensive, and lack reliability. Inspecting the large steel bridge inventory in the United States hence remains a great challenge due to the lack of a human-centered, efficient and cost-effective methodology for detecting, tracking, and documenting fatigue cracks. On the other hand, if crack inspections could inform the inspector in the field, more reliable, efficient, and accurate assessment of the inventory could be achieved and documented. Recently, computer vision has shown great potential as a non-contact, low-cost, and versatile platform for structural health monitoring (SHM). However, most computer-vision-based crack detection methods rely on still images to extract edge features of cracks. As a result, distinguishing real fatigue cracks from crack-like surface features, such as scratches, corrosion marks, and structural boundaries remains a major challenge. In addition, inspectors currently lack an effective way to efficiently interact with new and historic inspection data.  Such human-centered ability has been identified as one of the top interests of bridge inspectors, as it not only improves inspection quality but also facilitates decision-making in the field. To overcome the above challenges, this project proposed integrating computer-vision-based motion tracking and augmented reality (AR) techniques to empower bridge inspectors to perform robust fatigue crack detection, characterization, tracking, and documentation in the field. The developed computer vision algorithm does not rely on edge features of images. Instead, it is based on recording a short video of the structure under fatigue loading, tracking the surface motion through the proposed algorithm, and analyzing the surface motion pattern to reveal the ‘breathing’ of fatigue cracks. In addition, the crack width could be quantified with sub-millimeter accuracy using the tracked surface motion. To overcome the limitation of the technique in the field for the inspectors, this project research integrated computer vision with Augmented Reality (AR) to enable inspectors see crack information such as the crack geometry, realized via holograms overlaid on top of the bridge surface. The developed wearable AR device is expected to greatly increase bridge inspectors’ ability to perform accurate and reliable on-site inspection in a human-centered manner. Furthermore, inspectors will be able to interactively manage inspection results and compare with historic data for efficient decision-making.

The final report is available.