An Integrated Fire Scene Information System Based on YOLOv4-Tiny and DeepSort Technologies

This research introduces a pioneering framework for flame, smoke detection, and personnel monitoring, developed to address the increasing frequency of fire-related incidents in Taiwan. These incidents have caused significant loss of life and property damage. To tackle these issues, we employ YOLOv4-Tiny and DeepSort technologies to detect flames, smoke, and monitor personnel within surveillance footage, providing essential information to users. Unlike previous studies, this system integrates flame and smoke detection with personnel monitoring, enhancing rescue efficiency and ensuring personnel safety. During testing, the system effectively analyzed danger levels, monitored personnel, and provided necessary information. However, smoke can reduce the accuracy of personnel monitoring, so future research will aim to improve this issue and optimize the detection and identification of individuals moving between different footage. In summary, this investigation proposes a system that analyzes danger levels and monitors personnel in fire scenes. A survey of firefighters and security personnel concluded that this system helps improve rescue efficiency and reduce personnel loss due to fire incidents.