BIM 모델과의 정합을 통한 드론 기반 포인트클라우드에서 구조부재 추출 방안

초록

Point Cloud-based 3D data are increasingly used for construction management. However, previous structural element extraction studies have mainly relied on terrestrial LiDAR. Drone photogrammetry enables rapid and non-contact data acquisition over large areas, but the resulting Point Clouds can be affected by occlusion, shooting direction, temporary structures, stored materials, and adjacent structural members. In this study, a Point Cloud generated from an actual reinforced concrete construction site was registered with an IFC-based BIM model. Columns, beams, walls, and slabs were extracted using point-to-mesh distance filtering and elementspecific rule-based processing. Extraction performance was evaluated against manually constructed Ground Truth in representative regions of interest using Precision, Recall, and F1-score. The F1-scores were 94.64% for columns, 69.83% for slabs, 60.63% for walls, and 56.99% for beams. Columns showed the most stable performance, while beams showed relatively low point-level performance despite maintaining continuous longitudinal geometry. These results indicate that extraction performance varies by structural element and is influenced by occlusion, adjacent elements, stored materials, and Point Cloud incompleteness. Therefore, IFC-BIM geometry and element-specific processing conditions should be considered together for reliable structural element extraction from drone-based construction site Point Clouds.

키워드