Smart SSOMA: An Integrated Artificial Intelligence-Driven Automation System for SSOMA Risk Prevention in Industrial and Construction Environments

Abstract

This paper presents a comprehensive intelligent artificial vision system designed to automate risk prevention in industrial and construction environments by supporting occupational health and safety supervisors (SSOMA). The proposed solution integrates real-time camera-based monitoring with trained deep learning models capable of detecting unsafe behaviors such as improper use of personal protective equipment (PPE), incorrect manual load handling postures, and other hazardous actions. The system processes video streams locally using optimized computer vision algorithms and neural network models trained on labeled datasets to ensure reliable detection under varying environmental conditions. Upon identifying a safety violation, the system automatically generates visual alerts, stores photographic evidence, and logs the event in a structured database. All information is centralized in a local server developed using Flask, which provides an interactive web-based interface for real-time supervision, historical incident review, and risk analytics. The architecture prioritizes data privacy through on-premise deployment, low latency response, and scalability for multi-camera industrial scenarios. Experimental validation demonstrates high detection accuracy, rapid response times, and operational robustness, highlighting its potential as a practical, cost-effective, and scalable Industry 4.0 solution for proactive workplace risk management.