论文部分内容阅读
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 investigate the rapid and non-destructive testing methods of corn seed vigor based on near-infrared diffuse reflectometry. Methods In this study, the spectral curves of 450 ~ 900 nm of M017 maize seeds with no apparent damage on the surface were obtained by using the near-infrared spectroscopy system established by the laboratory. The calibration set and verification set ratio were 3: 1. Red ink staining method to determine whether the corn seed samples have vitality. The principal component analysis (PCA) was used to extract the main discriminant components by using the SG-5 pretreatment method to reduce the curve noise. Based on the measured seed vigor and its spectrum Curve is analyzed by using support vector machine (SVM) to establish discriminant model. Results When the cumulative contribution rate reached 96%, the six principal components were selected to establish the model with the highest correct discrimination rate. The near-infrared diffuse reflectance spectroscopy data could better discriminate the seed vigor, and the correctness of the calibration set and the predictive set Respectively 95.56% and 86.67%. The conclusion proves that this method is feasible and can basically meet the requirements of rapid nondestructive testing to determine the seed vigor of maize, which lays the foundation for the rapid and nondestructive detection of maize seed vigor in the future.