文章回顾了本研究小组将人工神经网络(ANN)和深度神经网络(DNN)用于提取布里渊光时域分析仪(BOTDA)传感系统中的温度分布信息的研究工作。在对ANN或者DNN模型进行适当的训练之后，沿被测光纤的温度分布信息能够被训练完成的ANN或DNN从实验获得的布里渊增益谱(BGS)中直接提取出来，而不需要像传统的洛伦兹线型拟合(LCF)方法一样先对BGS拟合得到布里渊频移(BFS)，再将其转换成温度信息。实验结果展示出了用ANN和DNN进行温度提取的方法相比于用传统LCF方法的优势。

This paper reviews the research work of Artificial Neural Network (ANN) and Deep Neural Network (DNN) for extracting temperature distribution information in Brillouin Optical Time Domain Analyzer (BOTDA) sensing system. After proper training on ANN or DNN model, the Brillouin gain along the measured temperature distribution of optical fiber can be extracted from the Brillouin Gain Spectrum (BGS), rather than the traditional Lorenz Curve Fitting (LCF) method which fits the BGS as Brillouin Frequency Shift (BFS), and then convert it to the temperature information. The experimental results show the advantages of using ANN and DNN for temperature extraction compared with the traditional LCF method.