The Performance of the Optical Flow Field based Dense Image Matching for UAV Imagery

We compiled an inventory of building condition (destroyed vs. survived) following the 2024 Noto Peninsula Earthquake for the entire affected area, totalling over 140,000 structures. We discuss how we fused freely available data of different types and from different sources to generate the final dataset. We show that our method produces highly accurate results relative to those obtained by on-site surveys. This data can be used to train AI to quickly detect damaged structures in future disasters.