基于近红外化学成分特征吸收光谱的野生台蘑鉴别模型研究

本文建出一种应用近红外光谱技术鉴别野生台蘑的新方法。使用FieldSpec3便携式近红外光谱仪对包括野生台蘑在内的13种蘑菇进行漫反射光谱采集。将采集的数据经小波去噪后,对可见与近红外光谱(350～2500nm)进行峰谷筛选,所得峰谷集经主成分分析降维,取方差贡献率大于99.9%的5个主成分作为BP神经网络的输入值,建立数学模型。该模型在偏差±0.05内,对未知样本正确识别率为100%。本结果表明利用近红外漫反射光谱可以很好地鉴别野生台蘑。

Study on Recognition Model for Wild Wu Tai Mushroom Based on the Chemical Composition of Near Infrared Spectral Diagnosis

A new method for discrimination of wild Wu Tai Mushroom by means of Near infrared spectroscopy (NIRS) was established.Visible and near infrared reflectance spectra of 13 different types of mushroom including 3 wild Wu Tai Mushrooms were collected with a portable Near infrared spectrometer (ASD Fieldspec3).Spectra peaks and vales were selected by peakFinder in Matlab after the spectral data was denoised by wavelet transform(WT).The first 5 principal components (5PCS) were computed by principle component analysis(PCA),and their accumulated variance contribution rate was more than 99.9%.Finally,a back propagation artificial neural network model(BP-ANNM) was built by use of the 5 PCS as inputs to the BP-ANNM.The discrimination rate of 100% was achieved for unknown samples.It suggested that wild Wu Tai Mushroom could be identified very well with NIRS.