Image-Based Inspection System of Firefighting Facilities Using Augmented Reality

Abstract

With the rapid development of digital-based industrial structures, the construction industry is also rapidly adopting smart-based technologies and unmanned automation, leading to an increase in the construction of new buildings. To protect against fire, disasters, and other emergency situations, installing firefighting facilities is mandatory , and continuous management is required. However, owing to the increase in new buildings, the installation of firefighting facilities is increasing, but the manpower to manage and inspect these firefighting facilities is lacking. In this paper, a technical improvement plan is proposed for the existing analog method of facility inspection. By combining deep learning algorithm and augmented reality, a fire protection inspection system is developed and inspections are performed through wearable devices. To train the object detection model, we built a dataset of five types of firefighting facilities. The trained YOLO v4 model achieved a performance accuracy of 93.6%. The proposed system is expected to perform more convenient and reliable inspections and to help with fire safety management by enabling both fire safety professionals and the general public to learn about fire protection inspection systems and perform self-inspections.