Abstract
Working at nuclear facilities subjects workers to a number of industrial health and safety risks. During their normal duties, workers are potentially exposed to hazardous processes and materials (e.g., hot steam, harsh chemicals, electricity, and pressurized fluids) the facilities may contain and other hazards (including slips, trips and falls). Nonetheless, even though workers are trained to stay away from potential dangers, there are still many types of risks that can occur within only a few minutes of carelessness. Occupational safety and health (OSH) monitoring at nuclear facilities requires observing and identifying a variety of specific unsafe behaviors with feedback to on-site workers. However, this mostly relies on manual observation and recording, which is time-consuming and costly. To this end, this paper presents an automated identification approach to enhance OSH management in nuclear facilities. First, a visual information extraction module integrating the state-of-the-art deep learningbased models is proposed to obtain hybrid visual information. Subsequently, on-site occupational hazards are identified by automatically analyzing relationships of detected entities. The experimental results demonstrate that with high-performance extraction and analysis of hybrid visual information, occupational hazards can be effectively and automatically identified for intelligent OSH management.
