Deformation measurement of twisted timber beam using UAV SfM photogrammetry and a new feature extraction algorithm

Abstract Drones facilitate the monitoring of large structures through feature extraction from point clouds generated through Structure-from-Motion photogrammetry. In the present study, we determined the deformation of a structural timber strip subjected to simultaneous bending and torsion. Three cameras were used. Two of them are pre-installed on the UAVs utilized, and the third is a consumer-grade Canon camera. All three were configured in flight mode. The geometry of the timber strip was generated through photogrammetry from the photos taken with each camera at a height of 1.5 m. The results were compared with the reference geometry, which was also created using the Canon camera on the ground at an average distance of 0.92 m. This reference geometry was previously validated in a preparatory project using extensometers with 1-µm precision. A Python-based algorithm was developed to automatically extract the position of the centroid and the rotation of each cross-sectional segment of the strip from UAV-based photogrammetric point clouds. Deformations measured by each of the three devices and the new algorithm are compared with actual deformation. The accuracy in measuring displacement and rotation of the centroid of strip cross-sections ranged between − 0.05 and 0.09 mm and between 0.00° and 0.24°, respectively.