以2009年产自海南的农华101玉米种子作为研究对象, 基于近红外漫透射光谱法(波长范围908.1~1 677.2 nm), 研究了一种高强高效近红外光源在不同光源电压以及光源至光谱仪不同距离两种条件下对玉米杂交种鉴别的影响。 对光谱进行一阶导数、 矢量归一化的预处理后, 使用主成分分析(PCA)和正交线性判别分析(OLDA)提取光谱特征, 使用支持向量机(SVM)分别建立种子纯度鉴定模型, 统计不同实验条件下的识别率。 结果表明, 在电压较低或者光源至光谱仪的距离较大的时候, 光源强度较低, 得到的光谱曲线有较多的毛刺, 此时的识别率较低, 增大电压或者降低光源至光谱仪的距离时, 光谱曲线变得较为平滑, 识别率明显升高, 说明在一定范围内增大光源强度会提高模型的正确鉴定率。

This paper focus on the effects of a high strength and high efficiency near infrared light source on the identification of corn hybrids under different light source voltage and different distances from the light source to the spectrograph based on the near infrared transmission(NIT) spectroscopy (wavelength range 908.1~1 677.2 nm) with the Nonghua 101 corn seeds harvested from Hainan province in 2009 as the research object. After the first order derivative and vector normalization of the spectra, the spectral characteristics are extracted using principal component analysis (PCA) and orthogonal linear discriminant analysis (OLDA) before the establishment of the model using support vector machine (SVM). Then the recognition rate under different experimental conditions is caculated. The results show that the lower voltage source or a larger distance from the spectrometer to the light source causes lower light intensity resulting to the spectrum curve with more burrs and the lower recognition rate. By increasing the voltage or decrease the distance from the light source to the spectrometer, the spectral curve becomes relatively smooth, and the recognition rate is significantly increased, indicating that the rate of correct identification of the model can be enhanced by increasing the light intensity within a certain range.