Image-Based 3D Reconstruction for Posthurricane Residential Building Damage Assessment

Street-level storm damage photos are an essential type of data used in postdisaster damage assessment. However, the existing approaches only leverage such data in a two-dimensional context. The research reported in this paper leveraged the photos collected during Hurricane Sandy to explore image-based three-dimensional (3D) reconstruction for posthurricane residential building damage assessment. Specifically, two commonly used image reconstruction pipelines are employed to reconstruct several impacted residential buildings to evaluate their performances regarding key measurement needs in posthurricane damage assessment. Damage data recorded by a mobile light detection and ranging (LIDAR) system were used as the ground truth for performance evaluation. The study results suggest that image-based 3D reconstruction can adequately support hurricane damage assessment needs for residential buildings. However, for damage assessment tasks which rely on very accurate estimate (generally <1 cm) of displacement, deflections, and other distresses, it requires more accurate methods than structure from motion (SFM)-based 3D reconstruction.