Computer vision-based monitoring method of non-wearing helmet events using face recognition

Abstract

Abstract Wearing helmets is crucial for ensuring the safety of workers in the construction industry because this is the first line of avoiding over 70% of production safety accidents. However, many workers are not willing to wear helmets due to discomfort and reduced work efficiency. To this end, this paper proposes a computer vision-based monitoring method using face recognition to detect and prevent non-wearing helmet events on construction sites. Compared to existing surveillance or monitoring systems, the proposed method has three significant advantages. Firstly, by using a unique structure, the proposed method can achieve up to 97.7% accuracy in detecting workers not wearing helmets. Secondly, the proposed method enables real-time detection, allowing it to prevent dangerous behaviors by stopping them in advance. Finally, the proposed method has been successfully deployed on over 20 real construction sites, and it has detected more than 18,000 related events.