Published in 2025
This research work focused on evaluating both the technical and design elements necessary to perform augmented reality (AR) assisted bridge inspections. The inspection prototype developed by this project aims to demonstrate the potential benefits and limitations of incorporating head-worn displays into the inspection workflow. By virtue of switching to an electronic data collection method, many common quality assurance issues can be mitigated. Furthermore, AR has an additional advantage over tablet-based solutions in its hands-free operation, which can allow for increased safety and flexibility. Furthermore, data collected in an AR system is georeferenced locally to the bridge and shown to be adequately accurate to allow navigation to previously documented defects during the inspection. Finally, and most importantly, AR serves as a central platform within which many up-and-coming digital tools (such as artificial intelligence) can be integrated. To better understand this potential, the work also considered the use of AI for a specific use case (concrete crack measurements).
A prototype application was developed using the Microsoft HoloLens 2 headset, but the principles employed could be translated to other similar mixed reality platforms. The prototype targeted efficient integration into the entire field inspection workflow. Three key functionalities were identified and addressed as main challenges to this holistic inspection approach: 1) a structured data entry interface, 2) georeferenced annotations for visualizing historical data, and 3) computer vision for automated defect labeling. A unique research focus was how to address these challenges in imperfect, real-world field scenarios involving both novice and professional bridge inspectors. To this end, the research team explored a range of AR interaction techniques of various degrees of automation, drawing on human-computer interaction principles. The study's evaluation metrics focused on measurement accuracy, time on task, as well as the tool's impact on perceived workload and usability.
Overall, the research found that performing virtual measurements within the AR environment resulted in similar usability as traditional measurements (i.e. tape measure and crack ruler), while significantly reducing the cognitive workload. Results were mixed in terms of time on task and accuracy. In general, more automated measurement processes significantly reduced time on task, but tended to decrease accuracy. One key finding was that the best combination of metrics was found using hybrid automation approaches, where the inspector was able to refine automated defect labels. In terms of added value, less cognitive load could mean less errors in documentation process, saving time at the office. This would be added to the time saving in report generation due to the virtue of digital data collection. The platform also provides the ability to collect much more granular data, which could help improve overall data quality.
The research team was encouraged by the reactions of a small set of VDOT bridge inspectors during the field validation of the final prototype. These inspectors were impressed by the technology's capabilities and intuitive workflow, voicing support for the future potential for bridge inspection. While the performed work supports the functionality of the platform over the entire field inspection workflow, the current prototype still requires additional development to be fully implementation ready.
Supplemental files can be found at https://library.vdot.virginia.gov/vtrc/supplements
Last updated: January 31, 2025