Real-Time Stereo Vision and RTK GPS-Based Mapping Framework for UAV Powerline Inspection

Powerline inspection is a critical maintenance task that was performed by helicopters or, more recently, by manually piloted drones, both without automation or any real-time awareness of data completeness. Autonomous UAVs with advanced sensors have the potential to overcome these issues, but real-time mapping and autonomous navigation in complex outdoor environments remain challenging. Fraunhofer Italia aims to address these challenges by developing a UAV and ground station system to support maintenance teams in real-time infrastructure inspection, with a focus on powerlines. This thesis, in collaboration with Fraunhofer Italia, proposes a perception framework for geo-referenced 3D point cloud mapping using stereo vision and high-precision GNSS. The resulting maps can serve as a basis for path planning and autonomous navigation. The system is developed and tested using a Pixhawk flight controller, Jetson Orin NX companion computer and ZED 2 stereo camera. While the Pixhawk provides GNSS and IMU data for pose information, the ZED 2 camera capabilities were exploited to handle the complex geometry of transmission towers, highlighting both its strengths and limitations. Built on the Robot Operating System (ROS), the framework is modular and interconnected, enabling real-time mapping performance. Due to limited onboard computational resources, further processing tasks such as path planning must be offloaded to a ground station. Since testing on a real drone was not feasible due to legal restrictions, a mock-up was built, demonstrating the potential of real-time UAV-based mapping of transmission towers.