Fast Intrusion Detection in High Voltage Zone of Electric Power Operations Based on YOLO and Homography Transformation Algorithm

Abstract

The high voltage zone of electric power is a typical high-risk work area. It is harm to the safety of operators and maintainers especially in the case of leakage and arcing. It is especially important for the electrical power safety to track the worker's inspection, monitor the operation behavior and follow the current position. The complicated environment limited the application of monitoring instruments. This paper proposes an operator positioning method in high voltage zone based on the CNN method. Firstly, a homography transformation algorithm is used to transfer the 2D and 3D scenes to enhance detection accuracy. Secondly, a graph sequence detecting method based on YOLO is used on the purpose of fast detection. Finally, a trajectory prediction and tracking method is proposed for the early warning. The onsite testing results show the precision and speed of proposed method compared with existing method.