以不同存储时间, 同一产地及收获时间的10个玉米品种种子为对象, 研究存储时间在玉米单籽粒近红外光谱真实性鉴定中, 对近红外光谱分析技术检测结果的影响。 利用1月份光谱数据建立品种真实性鉴定模型(单月建模), 分别鉴定2到12月的相同品种, 原始光谱采用平滑、 一阶差分和矢量归一化进行预处理, PLS-DA建立模型进行分析比较, 结果显示, 正确鉴定率均呈逐月下降的趋势, 同一品种的同一种子批, 由储藏开始建立的品种真实性鉴定模型已无法对储藏11个月后的该种子批进行高准确度的鉴定, 储存时间由1个月增加至11个月时, 模型的平均正确鉴别率降低26.27%, 这说明玉米种子的存储时间越长将降低应用近红外光谱鉴定品种真实性的鉴定准确度。 另外, 本研究发现玉米种子存储时间越长, 导致同一品种种子样品的光谱数据在空间分布上产生差异, 光谱数据离散化更明显, 重复性一致性越低, 使得玉米种子的真实性鉴定结果的准确性越低。 通过扩充建模集中易受干扰的信息的范围, 即将1年内在不同时间段里随取样时间变化而导致的在不同环境因素、 仪器因素及种子样品等变化因素下采集到的光谱数据, 均扩充到建模光谱数据中, 以增加根据扩充数据建立的近红外光谱预测模型的包容性。 通过1月与2月建模集联合后建立的包容性模型(联合建模), 之后分别对2016年3月—12月测试集的样品进行鉴定, 之后逐月增加建模集光谱数据, 并对非建模集所在月份进行逐月鉴定, 以京科968为例, 结果表明, 模型对建模集相邻月份的测试准确度较高, 之后逐月降低。 当建模集内加入1到6月份建模集内的特征光谱后, 包容性模型的平均正确鉴别率可稳定在92%以上。 通过以上方式, 对10个玉米品种进行了测试, 结果表明, 包容性模型对于玉米种子真实性的正确识定率相较于单月模型均有明显提高。 J92与XY211的平均正确鉴别率分别提高11.58%与7.71%。 将2016年整年的光谱数据均加入包容性模型的建模集中, 使测试集玉米杂交种2017年的平均正确鉴别率达到94.68%, 自交系达到95.03%, 为进一步研发专用模型和实用设备提供基础。

The study, targeting at 10 Maize varieties with different storage time and the same origin and harvest time, aims to study the effects of storage time on the results of the near infrared spectrum analysis technology applied in the near-infrared spectrum authenticity identification of maize single-kernel varieties. The authenticity model (monthly modeling) of breeds was established by using spectral data from January to identify the same samples which spectral data from February to December. The original spectrum was pre-processed by smoothing, first order difference and vector normalization. PLS-DA was used to establish the model for analysis and comparison, the results showed that the correct identification rate was decreasing month by month. The average correct identification rate of the model is reduced by 26.27% when the storage time is increasing from 1 month to 11 months, Which indicates that the longer the storage time of maize seeds is, the lower the accuracy of the near-infrared spectrum authenticity identification will be. This research also indicated that with the increase of the storage time of maize seeds, the spatial distribution of the spectral data of the same species but at different storage time is different. The discretization of spectral data becomes obvious, and the repeatability and consistency are reduced, which makes the accuracy of authenticity identification results of maize seeds is reduced. We endeavor to expand the models to centralize the range of the information that is easily interfered, that is, expand the spectral data collected under different environmental factors, instrumental factors and seed samples in different periods of time in 1 year to the modeling spectrum data to increase the inclusiveness of the prediction model of the near infrared spectrum based on the expanded data. Then, the inclusive model (joint modeling) has established by jointing the January and February modeling sets, after that, identifies the test set samples from March to December respectively, and then increases the model set spectrum data month by month, and the identifies the months that non-modeling set is located month by month. It taking JK968 as an example, the results showed that the accuracy of the model for the adjacent months of the modeling set is high, and then decreases month by month. When the feature spectrum of the model set is added from January to June, the average correct identification rate of the inclusive model can be more than 92%. In the above way, 10 maize varieties were tested, which can be seen that the correct identification rate of the inclusive model for maize seed authenticity is significantly higher than that of the single month model. The average correct identification rate of J92 and XY211 is increased by 11.58% and 7.71%, respectively. At the same time, in order to further improve the correct identification rate of the model, this study added the spectral data of the year 2016 to the modeling concentration of the inclusive model, so that the average accuracy identification rate of maize hybrids in 2017 reached 94.68%, and the inbred line reached 95.03%, providing the basis for further developing special models and practical equipment.