基于近红外光谱技术的三种硬实种子无损鉴定

摘要

To explore the applicability of the NIRS on nondestructive identification of hardseedness characteristics,in this study,near infrared spectroscopy (NIRS) models were established using near-infrared spectroscopy combined with Partial Least squares (DPLS) to investigate the hard seed characteristics of legume seeds,Codariocalyx motorius Glycine max Cassia tora L.and Sophora alopecuroides.120 seeds of the three species were divided into two groups:calibration set (80 samples) and validation set (40 samples),which contained the same number of hard seeds and soft seeds.The influences of replicate times,spectral region and calibration samples on the discrimination rate were compared.The results showed that with 4 000-5 000 cm-1 of spectral range,vector correction,eight main components,the predictive value of the model fitted with the true value well.Discrimination rates of calibration and validation sets in NIR models were over 85％ for Glycine max.With 4 000-8 000 cm-1 of spectral range,first-derivative spectroscopy,four main components,discrimination rates of calibration and validation sets were about 90％ for Cassia tora L.With 4 000-8 000 cm-1 of spectral range,second derivative spectroscopy,eight main components,discrimination rates of calibration and validation sets were over 95％ for Sophora alopecuroides.%为探讨近红外光谱技术在鉴定种子硬实特性上的普遍性,该文采用近红外光谱法结合偏最小二乘法建立了大豆、苦豆子和决明子单粒种子硬实特性的定性分析模型,每种种子均选择120粒种子进行近红外定性分析,种子分为建模集、检验集2组,建模集80粒,检验集40粒,各组中硬实与非硬实种子的比例为1∶1,比较了光谱重复次数、光谱范围以及不同建模样品的建模效果.结果表明:采用二次平均光谱所建模型的鉴别率优于单次光谱；大豆采用4 000～5 000 cm-1光谱范围,矢量校正预处理,主成分为8时,建模集与检验集鉴别率均在85％以上；决明子采用4 000～8 000 cm-1光谱范围,一阶导数预处理,主成分为4时,模型建模集与检验集鉴别率均在90％左右;苦豆子采用4 000～8 000 cm-1光谱范围,二阶导数预处理,主成分为8时,模型的建模集与检验集鉴别率均在95％以上.以上结果表明近红外光谱技术可以很好地应用于单粒种子硬实特性的判断,有助于硬实机理的深入研究.