Overview
To improve autonomy and safety in transmission facility inspection, a multimodal autonomous flight system using optical camera and 3D LiDAR is developed. The system perceives transmission towers and power lines using both visual and spatial information, and is designed to autonomously approach and track targets without relying on GPS or gimbal control. This enables precise and efficient inspection of transmission infrastructure even in complex environments.
Key contribution
▸ RoMP Transformer: Development of a transformer-based deep neural network using rotational bounding boxes and multi-level feature pyramids
▸ Autonomous tower approach method: Development of a method for autonomously approaching transmission towers without relying on GPS information
▸ Autonomous transmission line tracking method: Development of a method for stable autonomous tracking of power lines without gimbal control
Impact
▸ Advanced autonomous inspection capability: Establishment of a precise inspection framework applicable to diverse environments
▸ Enhanced safety and operational efficiency: Stable and efficient operation enabled by reducing dependency on external systems
▸ Improved deployability through lightweight system design: Enhanced field applicability through system simplification and lightweight UAV operation
▸ Reliable autonomous inspection in high-voltage environments: Enabling stable inspection performance in complex and high-risk transmission environments