MGLD-TLNet: Multigeometric and Long-Distance Representation Network for Transmission Line Inspection.

Area of Science:

• Electrical Engineering
• Computer Vision
• Robotics

Background:

• Effective transmission line (TL) inspection is crucial for reliable power delivery.
• Complex corridor environments pose challenges for current inspection methods, particularly with sparse and imbalanced point cloud data.

Purpose of the Study:

• To present a novel 3D-based perception method for TL inspection in challenging environments.
• To address data sparsity, imbalance, and structural correlations inherent in TL inspection tasks.

Main Methods:

• Modeling long-range spatial relations to mitigate data sparsity and imbalance.
• Constructing a unified 3-D representation that jointly models towers, conductors, and vegetation.
• Fusing Cartesian and polar geometries through geometry-aware alignment.

Main Results:

• The proposed method, MGLD-TLNet, demonstrates improved stability and accuracy.
• Consistent performance gains were observed under conditions of sparsity, occlusion, and complex environmental interactions.
• Experiments on real-world corridor datasets validate the effectiveness of the approach.

Conclusions:

• The developed 3D perception method enhances TL inspection capabilities in complex environments.
• Leveraging long-range spatial relations and fused geometric representations improves robustness and accuracy.
• This work contributes to more reliable power delivery through advanced inspection techniques.