Abstract

The rapid adoption of Industry 4.0 technologies has transformed manufacturing environments, integrating IoT devices, robotics, and automated production lines. While efficiency improved, ensuring worker safety in dynamic industrial settings remained a critical challenge. Traditional surveillance systems lacked real-time hazard detection and proactive intervention capabilities, leading to delayed responses during accidents or unsafe behaviors. Industrial accidents continued to occur due to inadequate monitoring of complex processes and the inability of conventional video surveillance to provide intelligent, context-aware safety insights. There was a pressing need for a system capable of detecting unsafe worker behaviors, equipment malfunctions, and environmental hazards in real time, with minimal latency and computational overhead. This study proposed an Edge-AI powered video analytics framework designed to operate directly on local manufacturing site devices. High-resolution video streams were preprocessed and analyzed using a lightweight convolutional neural network (CNN) model optimized for edge deployment. Object detection, motion tracking, and behavior classification algorithms were integrated to identify unsafe actions, equipment proximity violations, and hazardous zones. Alerts were generated in real time and transmitted to a central monitoring dashboard. The system was evaluated in a simulated smart manufacturing environment with multiple worker scenarios and equipment interactions. The proposed framework achieved a detection accuracy of 95.5% for unsafe worker actions and a precision of 94.3%, with a recall of 93.2% and an F1- score of 93.9%. Latency remained under 123 milliseconds per frame, enabling near real-time alerts.

Authors

G. Venkataramana Sagar¹, S. Ambigaipriya²
G Pulla Reddy Engineering College, India¹, Mookambigai College of Engineering, India²

Keywords

Edge-AI, Video Analytics, Industrial Safety, Smart Manufacturing, Real-Time Monitoring