Anomaly Detection of Power Transmission Line Using Stereo Vision-Based Multi-Rotor UAV

The distance between a line and the object below is an important index in the maintenance of power transmission, it must be controlled rigidly in tolerance. To accomplish the established task, we proposed a method of measuring the distance between line and the object below based on three-dimensional scene reconstruction using stereo vision on Unmanned Aerial Vehicle (UAV), including Multimodal Neural Networks (MNN)-based semantic segmentation on RGB-D data for power transmission line extraction, and a large-scale semantic scene reconstruction method based on voxel hashing and line modeling. Through fitted line models, we judge whether the distance conforms to safety condition or not. The experimental results demonstrate the practicability and accuracy of the distance measurement solution.