Substation Personnel Safety Detection Network Based on YOLOv4

With the development of more and more substations towards the direction of unattended, the substation in the field of personnel safety protection is becoming more and more automatic and intelligent. A video surveillance system is used to monitor the daily operating conditions of the substation. The application of more efficient video image analysis technology and detection technology to the monitoring system can effectively protect the substation staff in manual safety inspection. By more quickly detecting the movement track and position of personnel in the station, it is particularly important to give safety warnings to personnel who have entered the dangerous work area by mistake. Based on whether the substation personnel wear safety helmets as a benchmark, the article detects the wearing of the safety helmets of those entering high-risk areas to effectively distinguish between workers and non-workers. And through Yolo4's detection technology, managers can clearly monitor the situation in the dangerous area, realize the effect of predicting the occurrence of danger in advance and minimizing the danger. Therefore, in response to the above problems, this paper simplified the feature fusion part based on Yolov4 and achieved the target detection effect with a more lightweight feature extraction backbone network. The improvement of the above method has effectively improved the detection speed of the network and has strong robustness.