DMMP红外光谱的自动识别

Recognizing DMMP Infrared Spectra Automatically by PatternRecognition Techniques

介绍了模式识别技术在化学毒剂红外遥感监测领域应用的概况,探讨了线性分类器、分段线性分类器、反向传播人工神经网络(BP-ANN)分类器应用于红外光谱鉴别的可能性。用一个DMMP(甲基膦酸二甲酯)红外光谱数据样本集对上述三种分类器进行了实际的训练和鉴别性能预测,结果发现,分段线性分类器的性能优于另外两种分类器,鉴别率达到了80％以上。

The application of the pattern recognition techniques for remote infrared chemical detection was briefly introduced. The possibility that the linear discrimination analysis, piece-wise linear discrimination analysis, and back-propagation artificial neural network (BP-ANN) classifier implemented to discriminate infrared spectra was approached. A data set of DMMP (dimethylmethylphosphonate) infrared spectra was used to perform the actual training and discriminate rate prediction of these classifiers. The prediction results shows that the properties of piecewise linear classifier are better than other two classifiers, and the discrimination rates is more than 80%.