Application of near-infrared ( NIR) and attenuated total reflection Fourier transform infrared ( ATR-FTIR) spectroscopy technology in the detection of pesticide residue are studied. In order to analyze from different points of view, respectively, partial least squares ( PLS) and BP neural network are used to establish mathematical models. The NIR spectra is optimized by minimum-maximum normalized method before the model built. The standard deviation of NIR-PLS model, respectively,is RMSEC =0. 289,RMSEP =0. 335. The standard deviation of NIR-BP model is RMSEC =0. 436, RMSEP =0. 610. Besides, SNV method is used to preprocess ATR-FTIR spectra. Standard deviation of PLS model fop ATR-FTIR, respectively, is RMSEC =0. 168, RMSEP =0. 127. The standard deviation of BP model for ATR-FTIR is RMSEC =0. 100, RMSEP =0. 106. The results indicate that ATR-FTIR has a higher detection precision than NIR spectroscopy method in pesticide residues detection. ATR-FTIR spectroscopy has a real potential and advantage for future pesticide residues detection application.%研究了近红外(NIR)光谱技术和衰减全反射傅里叶变换红外(ATR-FTIR)光谱技术在萝卜农药残留检测中的应用.为了从不同角度分析,分别使用偏最小二乘(PLS)法和BP神经网络建立了数学模型.经最小-最大归一化方法优化后的NIR光谱数据的PLS法模型标准差为:RMSEC =0.289,RMSEP=0.335,BP神经网络模型标准差为:RMSEC =0.436,RMSEP =0.610.矢量归一化方法优化后的ATR-FTIR光谱数据的PLS法模型标准差为:RMSEC =0.168,RMSEP =0.127,BP神经网络模型标准差为:RMSEC=0.100,RMSEP =0.106.结果表明:ATR-FTIR的农药残留检测精度高于NIR光谱技术,ATR-FTIR光谱技术在农药残留检测方面有实际应用潜能和优势.
展开▼
