Efficient First Floor Height Estimation of Buildings from Sequential Images and Airborne LiDAR for Flood Risk Analysis

First Floor Height (FFH) is an essential parameter for flood risk analysis, where buildings with a lower FFH are at higher risk of content damage for a given base flood elevation. Despite its importance, FFH is often missing from building databases and can be time-consuming or expensive to collect. Here, an efficient and cost-effective method is proposed to accurately estimate FFH using high-resolution, sequential images extracted from video collected with a vehicle-mounted GoPro camera and used to derive a 3D representation of the environment. To improve positional accuracy, the 3D model is georeferenced using preexisting airborne LiDAR, achieving global map consistency. FFH is then determined as the difference in elevation between the door bottom extracted from GoPro data and the ground level determined from a surface terrain model aligned to a common global coordinate system. To improve efficiency, the Segment Anything Model for videos (SAM-2) is used to track door instances across image frames, facilitating FFH estimation. The proposed method achieves a Mean Absolute Error of 21 cm. Unlike other methods that estimate FFH from street-view imagery, our approach is more robust to occlusion by exploiting multiple viewpoints and does not rely on proprietary data sources, making it more reliable.