Reconstructing Power Lines from Images

We present a novel method for 3D reconstruction of overhead power lines from a few images. The solution to this problem has a deep impact over the strategies adopted to monitor the many thousand kilometres of powerlines innervating our countries;currently, the only effective solution is based on the use of high-end laser scanners mounted on drones. The difficulty with image-based 3D reconstruction algorithms is that images of wires of the power lines typically do not have point features to match among different images. We leverage on a few assumptions that can be made on the power lines case and define an ad-hoc strategy for solving the problem. We first use a Structure from Motion algorithm to retrieve the approximate camera poses and then formulate a minimization problem to simultaneously define a 3D wire hypothesis and refine the camera poses so that the projections of that wire are consistent on the supporting images.