open access

https://hdl.handle.net/2115/100471

UGV-Driven Indoor Damage Detection of Ceilings and Walls Using PatchCore

英語

wall surface

2026-07-01

Proceedings of the 10th International Symposium on Reliability Engineering and Risk Management (ISRERM 2026)

Post-earthquake safety assessment of building interiors requires rapid detection of damage to nonstructural elements such as ceilings and walls, which may affect occupant safety and building usability even after moderate earthquakes. However, the lack of damaged image datasets limits the application of supervised deep learning methods. This study proposes an indoor damage detection system that combines PatchCore-based anomaly detection with autonomous image acquisition using an Unmanned Ground Vehicle (UGV). The method detects moderate-scale ceiling damage and wall cracks using only normal-condition images. To enable efficient inspection, a LiDAR-based navigation algorithm is developed that allows the UGV to travel along walls without prior maps or GNSS. Experiments conducted in a real building environment, with ceiling damage simulated by opening inspection hatches, demonstrate that the UGV achieves stable autonomous navigation and that PatchCore effectively detects ceiling and wall damage. The proposed system enables rapid and scalable post-earthquake indoor safety assessment.

会議発表論文

Version of Record