基于红外热成像的隧道火焰检测技术研究

精准的火焰检测是有效避免火灾发生的关键，针对传统的火灾探测算法在公路隧道等大空间环境中存在及时性与准确性相互制约的问题，通过研究隧道火焰初期在图像中呈现的静态和动态特征，提出了一种基于红外热成像的公路隧道火灾初期火焰检测方法。利用温度阈值获取疑似火焰区域，根据红外图像在引导滤波器作用下降噪，同时利用区域增长法分割疑似火焰区域；从疑似区域中提取的特征值构成特征向量，进行数据归一化提高SVM收敛速度；利用人工蜂群算法优化参数。结果表明：ABC-SVM能够实现公路隧道火灾初期的火焰识别，检测正确率相较于RBF方法提升了2.26%,运行时间缩短了2.29 ms;检测正确率相较于SVM方法提升了0.87%,运行时间缩短了2.22 ms。本方法可以对初期隧道火灾进行快速、有效检测，并有良好的环境适用性。

Research on tunnel flame detection technology based on thermal infrared imaging

Precise flame detection is the key to effectively avoiding fire. This paper addresses the timeliness and accuracy problems in the traditional fire detection algorithm in a large space environment. An early flame detection method for highway tunnel fire is proposed by the infrared thermal imaging method based on the static and dynamic characteristics of the flame images. The temperature threshold is used to obtain the suspected flame area, and the noise is reduced according to the infrared image via the guide filter. The region growth method is used to segment the suspected flame area. The eigenvalues extracted from the suspected regions are used to form the eigenvectors to normalize the data and improve the convergence speed of SVM. An artificial colony algorithm is used to optimize the parameters. The results show that ABC-SVM can realize the initial fire identification of highway tunnel environment. Compared with the RBF and SVM detection models, the recognition accuracy is increased by up to 2.26%, and the running time is reduced by up to 2.29 ms.