Autonomous UAV navigation for close-range image acquisition of scaffold joints

Scaffold joints are critical structural elements that require continuous inspection. However, their small size and wide distribution make manual inspection inefficient. To address this challenge, this paper proposes an autonomous uncrewed aerial vehicle (UAV)-based method to efficiently capture close-range images of joints for inspection. The first module generates a real-time three-dimensional (3D) point cloud map using LiDAR and simultaneous localization and mapping (SLAM) and converts it into a bird's-eye view (BEV) image to provide spatial context. The second module plans scanning paths based on these data and performs three functions: (1) detecting scaffolds in the BEV image using a deep learning model and segments them into scan units; (2) evaluating scan feasibility based on UAV accessibility and obstacles; and (3) prioritizing and selecting the next target based on path efficiency and generates a collision-free trajectory. Experimental results at two outdoor construction sites demonstrated that the proposed method captured close-range images of all joints via autonomous flight, achieving 100% (207 joints) coverage. This study proposes a novel UAV-based method for autonomous close-range scanning of scaffolds, enabling precise monitoring of detailed components such as joints.