Lightweight Vision Architecture Deployed in Terminal for Safety Monitoring and Early Warning of Transmission Lines – Nature

A new lightweight vision architecture has been deployed on terminal AI platforms for safety monitoring and early warning of transmission lines, addressing critical gaps in real-time three-dimensional perception for power infrastructure. The system integrates pose estimation, visual detection, and depth transformation to enable precise ranging and hazard detection across complex multi-terrain scenes.

The architecture was developed to overcome limitations of current monitoring systems that rely on single-sensing modalities such as monocular/binocular vision or Light Detection and Ranging, which fail to achieve reliable real-time 3D perception and lack correlative analysis between external hazard intrusions and safe clearance distances of transmission lines.

By fusing pose estimation, visual detection, and depth transformation, the framework achieves high-precision ranging and early warning of external hazards within transmission corridors. The system employs a lightweight transmission line hazard detection model enhanced by a positive-negative sample dynamic balancing mechanism to improve detection performance.

An improved pose estimation algorithm enables high-precision spatial mapping, which, when combined with depth transformation and point cloud reconstruction, allows refined ranging for hazards relative to transmission lines under arbitrary terrain conditions.

The proposed method has been validated on terminal AI platforms and deployed on on-site camera terminals along transmission lines, demonstrating excellent inference performance and deployment adaptability on resource-constrained devices.

The research, published in Nature Electronics, contributes to advancing energy security and power distribution resilience through AI-driven monitoring solutions tailored for edge deployment in critical infrastructure environments.