Automatic autonomous vision-based power line inspection: A review of current status and the potential role of deep learning

Reviews automation of visual inspection of transmission and distribution lines and proposes a concept centred on UAVs, optical imagery and deep learning. Foot patrols and helicopter inspection are slow, costly and hazardous, while earlier automated helicopters and line-crawling robots had not been widely adopted because of strict accuracy demands. It discusses deep vision for both UAV navigation and inspection, with challenges and possible solutions.

Key findings

• Proposes UAV-based inspection with optical images as primary data and deep learning for analysis
• Identifies deep vision as useful for both navigation along lines and component inspection

Limitations

• Abstract only reviewed
• Conceptual proposal; no field validation in the review itself