Bridge assessment after 2024 Noto Peninsula earthquake and rainfall using point cloud

This study addresses the challenge of assessing bridge damage progression under compound disasters, specifically focusing on Bridge A in Japan affected by the 2024 Noto Peninsula Earthquake and subsequent heavy rainfall. We aimed to quantitatively evaluate potential changes in pier inclination using multi-temporal point cloud data (PCD) acquired by way of terrestrial LiDAR immediately after the earthquake and following the rainfall. Accurate pier isolation from complex PCD, complicated by environmental changes such as sediment deposition, was achieved using a deep learning semantic segmentation model enhanced with low-rank adaptation fine-tuning. Two distinct geometric analysis methods, M-estimator sample consensus (MSAC) cylinder fitting and principal component analysis (PCA) based on local surface normals, were independently applied to the segmented pier data to calculate inclination angles. Both MSAC and PCA consistently revealed a slight pier inclination ranging from 1.7° to 3° after the earthquake but detected no statistically significant progression after the heavy rainfall event. Our findings indicate no measurable progression of pier inclination during the observation period, demonstrating the effectiveness of the proposed workflow combining advanced scanning, segmentation, and robust multiple geometric analysis methods for monitoring structural stability and resilience in multi-hazard scenarios.