基于可见近红外光谱玉米种子活力的无损检测方法

Discrimination of vitality of maize seeds based on near visible infrared spectroscopy

目的：基于近红外漫反射技术,初步探讨玉米种子活力的快速、无损检测方法。方法本研究利用实验室自行搭建的近红外光谱检测系统获取60粒表面平整无明显损伤的M017玉米种子450~900 nm光谱曲线,其中校正集和验证集比例为3：1。利用红墨水染色法判定玉米种子样品是否具有活力。通过进行 SG-5点平滑（Savitzky-Golay smoothing, SG）预处理方法减小曲线噪声,基于主成分分析（principal component analysis, PCA）方法提取主要判别成分,并依据测定的种子活力情况和其光谱曲线应用支持向量机（support vector machine, SVM）建立判别模型进行分析。结果当累计贡献率达到96%时,选取6个主成分,建立的模型判别正确率最高,近红外漫反射光谱数据能够较好的判别种子活力的有无,其中校正集和预测集判别正确率分别为95.56%和86.67%。结论证明该方法可行,基本能够满足快速无损检测判别玉米种子活力的要求,为今后快速无损检测玉米种子活力奠定了基础。

Objective To explore a rapid and nondestructive method for the determination of seed vitality of maize seeds based on visible near infrared spectroscopy. Methods A laboratory visible near infrared spectroscopy system was built to collect reflectance spectra in the 450~900 nm of 60 M017 maize seeds with relative smooth surface and no damage, and the calibration set and validation set ratio of about 3：1. Then the red ink staining method was used to determine whether the maize seeds samples were with vitality. The noise of curve was decreased by the Savitzky-Golay （SG） pretreatment method and the main criterion constituents were extracted based on principal component analysis method （PCA）. Finally the discriminant model was established for spectral curve of seeds with different vitality based on support vector machine （SVM）. Results When the cumulative contribution rate reached 96%, 6 principal components were selected, and the highest discriminant accuracy of the model was set up. The near infrared diffuse reflection spectrum data could discriminate seed vitality, and the discriminant accuracy of calibration set and validation set were＆nbsp;95.56%and 88.67%, respectively. Conclusion The method is high reliability, and can satisfy the requirements of the germplasm resources and nondestructive testing seeds vitality, which laid a solid foundation for rapid nondestructive testing of maize seed vitality in the future.