Year: 2026
Type: article
Abstract: Facing the dual carbon goal and the demands of power grid intelligent upgrading, UAV-based inspection technology has been widely adopted for power equipment operation and maintenance. However, restricted by UAV load capacity and the complexity of insulator defect detection, existing methods struggle to balance real-time performance and reliable inspection for small defective targets. We improved the recognition performance of small defective targets by embedding an efficient multi-scale attention mechanism. A lightweight C3Ghost-based feature fusion layer was constructed to eliminate redundant computational overhead. For the backbone network, a progressive channel compression strategy was designed to cut network parameters. Experimental results demonstrate that compared with the YOLO11n baseline, the proposed ECP-YOLO11n reduces the parameter count by 59.7% and computational cost by 23.8%. Meanwhile, it retains 99.5% and 95.1% of the original model’s mAP50 on the CPLID-2000 and IDID datasets respectively, satisfying the requirements of real-time and high-reliability detection.
Source: Scientific Reports
Institution: Zhengzhou University of Light Industry
Language: English
Topic: Power Line Inspection Robots
Subfield: Mechanical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7229457970
