Research on Deep Learning-Based Recognition Technology for Violations in Live Electricity Operations

Abstract

Safety management and control in live electricity operation sites constitute a crucial assurance component for electrical safety production. As the demand for live electricity operations continues to rise, accompanied by increased complexity and difficulty, the shift from manual video analysis to intelligent control methods in on-site safety management has become imperative. In response to this, a human body posture recognition technology is proposed, utilizing YOLOv8 to establish a multi-person posture recognition model. This, combined with traditional image recognition techniques, achieves comprehensive perception of personnel states, enabling real-time management and early warning of hazards and non-standard behaviors during operations. This approach alleviates the pressure on inspection personnel and enhances the intelligence of violation recognition in live electricity operation sites.