基于YOLOv5网络模型的火焰检测

Flame detection based on YOLOv5 network model

煤炭资源在开采的过程中会伴随着产生一种名为煤层气的产物,煤层气又称为瓦斯,若将瓦斯直接排放至大气中,则会造成严重的温室效应,研究发现瓦斯可以通过燃烧用来发电,但瓦斯在发电过程中遇到明火,则会发生爆炸,给工作人员及企业会造成不可估量的损失,因此检测瓦斯发电站内的火焰情况,成为了解决瓦斯发电站爆炸事故的重要目标。基于火焰识别问题,采用传统目标检测算法难以满足精度要求,后续出现了基于深度学习的双步目标检测算法,虽在识别精度上能有效满足要求,但检测的实时性上存在不足。本文综合考虑目标检测的检测精度以及检测实时性,采用了最新的YOLOv5模型进行火焰的实时检测。

In the process of exploitation of coal resources can produce a product called coalbed methane,with gas,coalbed methane is also known as if gas direct emissions to the atmosphere,will cause serious greenhouse effect,the study found that the gas can be used to generate electricity by burning,but the gas power generation in flame,will be an explosion,to the staff and enterprise can create an immeasurable loss,Therefore,the detection of the flame in the gas power station has become an important target to solve the gas power station explosion accident.Based on the flame identification problem,it is difficult to meet the accuracy requirements of the traditional target detection algorithm.The subsequent two-step target detection algorithm based on deep learning appears.Although it can effectively meet the requirements in recognition accuracy,it has shortcomings in real-time detection.In this paper,the latest YOLOv5 model is used to detect flame in real time.