Cable detection and position estimation in a camera-based drone guidance system for autonomous sensor node attachment on transmission lines

With the shift towards volatile, renewable energy resources, monitoring high-voltage power transmission lines becomes essential for optimizing load and ensuring safety. The ASTROSE-System developed by Fraunhofer enables decentral-ized monitoring through self-sufficient sensor nodes that are directly attached to the power transmission lines and measure line tilt, line torsion, and current. However, the installation of the sensor nodes is a very ex-pensive process, as it currently requires trained workers, special equipment, and a power shutdown of the corresponding cable. An autonomous installation via drones would help, but without a power shutdown strong electromagnetic fields around the transmission lines disturb the operation of a GPS- or manually-controlled drone system. In this work, we propose a computer vision algorithm for a camera-based drone guidance system that allows the drone to orient itself and navigate relative to the cable positions. The algorithm consists of cable detection, cable extraction, and 3D stereo-matching for positioning and orientation measurements of transmission lines. The cable detection algorithm deploys a Gaussian second derivative filter to find the center of the cables. Only the detected centerlines are considered afterward in a robust and simple 3D stereo-matching algorithm, hence avoiding the complexity of the general problem. We then track the cables over time and predict the drone position relative to the cables with a simplified motion model and a Kalman filter. The algorithm and position and orientation measurements are implemented in Python and tested in a simulation environment using Blender.