Artificial Intelligence and Data Science
Online ISSN : 2435-9262
Effectiveness and challenges of acquiring crack damage inspection data using nighttime drone inspection-a demonstration experiment on Kakamigahara Bridge-
Katsuharu HIRATE, Kazuhide SAWADA, Keisuke YOSHIDA, Satoshi NISHIYAMA
Author information
JOURNAL OPEN ACCESS

2026 Volume 7 Issue 3 Pages 965-973

Details
Abstract

In recent years, expectations for drone-based bridge inspections have grown amid the aging of bridge infrastructure and a shortage of skilled engineers. In this study, a nighttime drone inspection was conducted on the Kakamigahara Ohashi Bridge to examine the effectiveness and challenges of acquiring data that can be utilized for routine bridge inspections. Even at night, images of the bridge underside and piers were successfully captured, and orthophotos and three-dimensional point clouds were generated. A comparison with the design values recorded in the bridge ledger confirmed that the acquired data can be used as practical supplementary materials. In the visual inspection of images, cracks with a known width of approximately 0.3 mm were identifiable, whereas the detection rate for finer cracks of approximately 0.2 mm remained at 18.2%. In addition, commercially available AI models produced false detections of graffiti, highlighting the need for standardization of imaging conditions and input image quality. Although on-site work time and direct labor costs were reduced, office work time increased, indicating that information-processing tasks constitute the primary workload.

Content from these authors
© 2026 Japan Society of Civil Engineers
Previous article Next article
feedback
Top