Smart SSOMA: Design and implementation of an adaptable, comprehensive SSOMA safety management system to reduce reporting times for PPE non-compliance, using machine vision and Edge Computing, applied to industrial and construction environments

Segundo J.; Garay J.; Chávez R.; Rodríguez F.; Castro F.; Antayhua R.; Espinoza J. · Proceedings of the LACCEI international Multi-conference for Engineering, Education and Technology (2026) · DOI: 10.18687/laccei2026.1.1.2671

Resumen

Manual supervision of Personal Protective Equipment (PPE) in Peruvian manufacturing SMEs presents critical limitations: discontinuous monitoring, human dependency, low traceability, and delays in SSOMA report generation, increasing the risk of occupational accidents and regulatory non-compliance. This research addresses the design and implementation of a mechatronic safety management system based on computer vision with YOLOv8 and Edge Computing, aimed at automating the detection of PPE non-compliance and the generation of SSOMA reports aligned with R.M. 050-2013-TR in the Peruvian manufacturing sector. The system integrates real-time video capture via IP cameras using RTSP/ONVIF protocols, image preprocessing, local inference with a custom-trained YOLOv8 model for helmet and reflective vest detection, automated event registration in a PostgreSQL database, and automatic PDF report generation through a FastAPI backend and web dashboard. All processing is executed locally on an Edge device, eliminating cloud dependency and ensuring low-latency response. Experimental evaluation over an 8-hour shift monitoring 35 workers achieved 94.6% detection precision, 92.8% recall, 93.7% F1-score, and an average inference time of 38 ms per frame with 82 ms total system latency. Automated report generation was completed in 40 seconds versus 30 minutes manually, representing a 97.8% reduction. User satisfaction reached 4.7/5. Results confirm that integrating YOLOv8 and Edge Computing into a mechatronic system delivers a viable, efficient, and scalable solution for occupational safety management in resource-constrained industrial environments.

Instituciones: Universidad Nacional de Ingeniería