Article Preview
TopBackground
Fire threatens human life and property safety, and in severe cases, it can cause huge economic losses and casualties. Nowadays, with the rapid development of the economy and the increasing scale of urban buildings, the fire situation has become more complex, and the difficulty of firefighting also increases. Currently, the main method of firefighting is still manual firefighting by firefighters, which is often accompanied by injuries or even sacrifices of firefighters. Therefore, developing more advanced firefighting methods and using robots to replace manual firefighting has become a research trend. Fire robots mainly use cameras to explore the situation of the fire scene and accurately recognize flame targets based on video images (Chen, 2023; Liang et al., 2024). This is the key to efficient robotic firefighting, and fire robots have significant real-time and accuracy requirements for flame target detection. With the increasing maturity of unmanned aerial vehicle (UAV) technology and the further expansion of aerial photography technology, UAV is increasingly being used in large-scale rescue equipment and intelligent fire detection (Chen, 2022; Wang, 2017).
Purpose of the Study
This article combines UAV platforms, airborne binocular vision, airborne processing computers, development workstations, and visual navigation integrated development environments. Deep learning systems (Kim & Muminov, 2023) can identify fire points quickly, and fire hazard inspections, on-site rescue command, fire detection, and prevention and control can be carried out on complex terrain and structural buildings in the air (Li, 2023). These methods solve the problems of traditional fire detection methods and improve the efficiency and accuracy of fire detection.
Research Status of UAV Fire Detection
In the field of UAV fire detection, researchers usually use image processing and machine learning techniques to achieve fire recognition. Among them, the methods based on image processing include extracting color features, texture features, and shape features, and motion detection based on optical flow and inter-frame difference techniques. Machine learning methods include traditional support vector machines (SVM), random forests, and deep learning methods such as convolutional neural networks (CNNs). In recent years, deep learning-based methods have achieved remarkable results in fire recognition. Among them, the you only look once v5 (YOLOv5) algorithm is an efficient object detection algorithm suitable for processing large-scale data sets and achieving real-time object detection. The YOLOv5 algorithm combines the characteristics of fast, accurate, simple, and lightweight and has high computational efficiency and recognition accuracy.
In UAV fire detection, researchers can use the YOLOv5 algorithm for fire recognition. By training the model to use the fire image data set taken by the UAV, the rapid detection and location of the fire area can be realized, and disaster relief measures can be taken in time. In addition, the YOLOv5 algorithm can combine other sensor data, such as infrared images and smoke sensor data, to improve the accuracy and reliability of fire detection. In general, fire recognition methods based on image processing and machine learning have achieved certain results in the field of UAV fire detection, and YOLOv5 algorithm, as an efficient object detection algorithm, is expected to provide more possibilities for the further development of fire detection technology.