Uncertainty-guided 5-DOF UAS path planning optimization for photogrammetric 3D bridge reconstruction

Abstract

Unmanned Aerial Systems (UAS) offer an efficient solution for photogrammetric 3D reconstruction and inspection of infrastructure, but planning optimal flight paths for complex structures such as bridges remains a significant challenge. This paper presents a theory-driven path planning framework that directly derives triangulation uncertainty from Structure-from-Motion (SfM) theory and incorporates it as a penalty-based optimization objective. This formulation enables dynamic allocation of viewpoints and promotes more uniform and reliable reconstruction quality. To further enhance robustness, an explicit overlap constraint is introduced to improve camera registration. Unlike prior studies that formulate the problem on offset surfaces, the proposed method defines path planning in a five-degree-of-freedom (5-DOF) search space, allowing flexible planning in complex 3D environments. Validation with SfM using images rendered in a virtual bridge environment demonstrates that the proposed approach generates point clouds with fewer missing regions, higher reconstruction accuracy, reduced misalignment issues, and improved efficiency. These results highlight the advantages of integrating theory-based reconstruction principles into 5-DOF photogrammetric path planning to achieve high-quality, reliable, and efficient UAS-based reconstruction of complex infrastructure.