Efficient visual inspection of fire safety equipment in buildings

Abstract

Fire safety equipment (FSE) in buildings is critical in ensuring occupant safety and mitigating losses during emergencies. However, its effectiveness is frequently compromised by inadequate maintenance. As buildings increase size and complexity, traditional manual inspection methods become impractical due to scalability and data management challenges. To address these issues, this paper proposes an advanced FSE detection framework with improvement strategies. The process commences with the developed YOLO-FSE algorithm, which is capable of identifying objects of varying sizes. This is complemented by precise localization of these objects through an enhanced tracking algorithm and visual simultaneous localization and mapping (vSLAM). The experiments demonstrate that this approach can effectively detect various fire safety equipment with the potential to replace labor-intensive manual methods. Notably, the YOLO-FSE network achieves a 7.9 % improvement in mean Average Precision (mAP) at a threshold of 0.5 (mAP@0.5), and a 9.4 % increase in mAP@0.95, indicating significant enhancements in detection accuracy.