Fast 3D Reconstruction of Power Transmission Lines Based on Sequential Images

The 3D reconstruction can visualize the inspection scene of power transmission lines (PTLs), providing data support for PTL intelligent inspection. To address the low-efficiency and large-computation of the 3D reconstruction methods for PTL, this paper proposes an efficient reconstruction method based on the sequential images of PTLs. Firstly, collect multi-view sequential images of PTLs by the Flying-walking Power Line Inspection Robot (FPLIR). Secondly, obtain the initial sparse point clouds data by the Structure from Motion (SFM) algorithm. Thirdly, remove the redundant point clouds based on the camera trajectories to acquire the point clouds data of PTLs. Lastly, obtain the dense point clouds of PTLs data and spatial surface model by the Multiple View Stereo (MVS) algorithm. The experimental results show that our proposed method allows processing 70 images within 48 minutes, obtaining the PTL model with a mean reprojection error of about 1.2 pixels. The proposed method provides the foundation for a refined and intelligent inspection of PTL.