Year: 2025
Type: article
Abstract: Industrial safety is a critical factor determining the sustainability and productivity of on-site operations. Although Information and Communications Technology (ICT)-based industrial safety technologies have been actively developed recently, the digital transformation of industrial sites is still constrained by insufficient infrastructure and high costs. This study designs and implements an Appropriate Safe Factory Module based on Vision AI, utilizing low-cost edge computing and Open Source Software (OSS) to promote the dissemination of ICT-based smart industrial safety technology. The target process is the pipeline welding process for petrochemical facilities, and the module aims to provide advanced safety management capabilities by simultaneously detecting non-wearing of Personal Protective Equipment (PPE) (safety vests and helmets) and the high-risk posture of falling. The proposed module consists of a Raspberry Pi 5, a representative small-scale edge computer, and an easy-to-install dome-shaped webcam, constructed for $120. Furthermore, optimization techniques, such as conversion to the Open Neural Network Exchange (ONNX) format, were applied during development to maintain real-time detection performance (1–5 fps) even in a low-spec computing environment. The research results confirm that the proposed module can be used as an appropriate smart industrial safety technology that enhances technological acceptability and security through on-premises, on-device inference in small and medium-sized enterprise (SME) environments, thereby achieving a practical improvement in safety management.
Institutions: Konkuk University, Korea Military Academy
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7117145192
FWCI:
Cites:
Cited by:
Related to: