Attention-Guided Autoencoder Fusion for Insulator Defect Detection Using UAV Transmission-Line Imaging
arXiv:2606.06536v1 Announce Type: cross Abstract: Automated defect detection in high-voltage transmission-line insulators remains challenging due to severe class imbalance, large scale variation, and the small spatial extent of defect instances in Unmanned Aerial Vehicle (UAV) imagery. To address these challenges, this paper proposes AE-YOLO, an Attention-Guided AutoEncoder-Enhanced YOLO framework for robust insulator defect detection. The architecture integrates lightweight bottleneck autoencoders within a Feature Pyramid Network-Path Aggregation Network (FPN-PAN) neck. This preserves anomaly
The proliferation of UAV technology combined with advancements in computer vision and AI models like YOLO makes efficient infrastructure inspection feasible at a critical time for grid reliability.
Improving the robustness and automation of critical infrastructure inspection directly impacts grid stability, operational costs, and the resilience of energy systems.
This development enhances the efficiency and accuracy of identifying defects in high-voltage transmission lines, reducing manual inspection needs and potential failures.
- · Energy utilities
- · UAV inspection service providers
- · AI computer vision developers
- · Infrastructure maintenance sector
- · Traditional manual inspection crews (some tasks)
- · Companies with less advanced inspection technologies
Automated drone inspections become standard for transmission line maintenance, improving reliability and reducing costs.
Reduced infrastructure failures lead to fewer power outages, enhancing energy security and supporting continuous operations for compute-intensive industries.
The success of this approach could accelerate similar AI-driven inspection and maintenance solutions across other critical linear infrastructure (e.g., pipelines, railways), increasing overall system robustness.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG