基于小波包能量分析的红外火焰信号识别

发布时间：2018-08-31 17:48

【摘要】：在设计红外火焰探测器的过程中,人工光源常常会引起探测器的误报。为有效区分人工光源与火焰信号,本文首先对采集的1种人工光源以及3种火焰信号进行分析,将采集的信号进行小波包4层分解,得到信号的能量谱。通过分析发现选取第2,3,4,7频段的能量值能够将这4种信号有效区分。为进一步验证实验结果,本文将双通道两路信号的第2,3,4,7频段的8个能量值作为一组特征向量,与BP神经网络结合进行模式识别。结果表明,通过这样的方法不仅可以区分火焰和人工光源,同时可以对3种火焰进行识别,其识别的正确率为84.1%。因此,基于小波包能量分析的方法提取这8个能量值作为特征值具有一定的可行性,能有效减少人工光源引起的误报,同时为火焰种类的识别以及以后的灭火自动化提供了新的可能性。
[Abstract]:In the process of designing infrared flame detector, artificial light source often causes false alarm of detector. In order to effectively distinguish the artificial light source from the flame signal, this paper firstly analyzes one artificial light source and three kinds of flame signals, and decomposes the collected signals into four layers of wavelet packet to obtain the energy spectrum of the signals. It is found that the four signals can be effectively distinguished by selecting the energy value of the second band. In order to further verify the experimental results, eight energy values of the second band of the two-channel and two-channel signals are taken as a set of eigenvectors, which are combined with BP neural network for pattern recognition. The results show that this method can not only distinguish the flame from the artificial light source, but also recognize the three kinds of flame. The recognition accuracy is 84.1%. Therefore, it is feasible to extract these eight energy values as eigenvalues based on wavelet packet energy analysis, which can effectively reduce the false positives caused by artificial light sources. At the same time, it provides a new possibility for the identification of flame types and the automation of fire extinguishing in the future.
【作者单位】：江南大学物联网工程学院;无锡格林通安全装备有限公司;
【分类号】：X932;TP183

【相似文献】