甜瓜的品种多样, 富含多种营养成分, 甜瓜种子品种不纯将对甜瓜生产造成一定危害, 研究采用种子的叶绿素荧光光谱结合反射光谱的分析方法鉴别甜瓜种子品种, 以甜瓜品种“一特白”、 “一特金”、 “京蜜7号”、 “京蜜11号”、 “伊丽莎白”为研究对象。 构建了甜瓜种子品种鉴别光谱系统, 包括激发光源单元、 光谱数据采集单元和数据处理单元, 使用该系统获取不同品种甜瓜种子的光谱数据。 对光谱数据分别进行一阶导数(first derivative, FD), Savitzky-Golay(SG) 平滑, FD结合SG平滑预处理。 采用主成分分析(principal component analysis, PCA)方法降低光谱数据的维数, 提取主成分。 使用两种不同分组方法将样品按照3∶1的比例分为训练集和验证集, 并分别采用Fisher判别和Bayes判别分析方法建立甜瓜种子品种的判别模型。 本文比较了仅使用叶绿素荧光光谱与使用叶绿素荧光光谱结合反射光谱建立判别模型的判别结果, 结果显示, 使用叶绿素荧光光谱结合反射光谱建模的判别结果优于仅使用叶绿素荧光光谱建模的判别结果, Fisher判别分析和Bayes判别分析的验证集样品品种的判别正确率均达到98.0%。 研究结果表明, 采用叶绿素荧光光谱结合反射光谱鉴别甜瓜种子品种具有可行性。

Melon is popular with us for its high nutritional value, and there are many varieties of melons. Impurity of melon seed variety will cause harm to melon production. This research adopted chlorophyll fluorescence spectrum combined with reflectance spectrum to identify melon seeds variety. Seeds whose varieties were “Yi Te Bai”, “Yi Te Jin”, “Jing Mi No.7”, “Jing Mi No.11”, “Yi Li Sha Bai” were used as research samples. A melon seeds variety identification system based on spectrum technology was developed, and it included an excitation light source unit, a spectral data acquisition unit and a data processing unit. This system was used to obtain fluorescence spectrums and reflectance spectrums of different varieties of melon seeds. First derivative (FD), Savitzky-Golay (SG), and FD associated with SG were utilized to preprocess spectral data respectively. Principal component analysis (PCA) method was adopted to reduce the dimensions of spectral data and extract principal components. This study adopted two different grouping methods to divided samples into training set and validation set according to the proportion of 3∶1, and Fisher discriminant analysis and Bayes discriminant analysis methods were used to establish discriminant models of melon seeds variety respectively. This study compared the discriminant effect of the model developed only using chlorophyll fluorescence spectral information with the discriminant effect of model developed based on chlorophyll fluorescence spectral information combined with reflectance spectral information. Results showed that discriminant model developed using chlorophyll fluorescence spectral information combined with reflectance spectral information generated better determination results than only using chlorophyll fluorescence spectral information, and the discriminant accuracies of validation set reached 98% in both Fisher discrimination analysis and Bayes discriminant analysis. Research results showed that chlorophyll fluorescence spectrum combined with reflectance spectrum technique was feasible for melon seeds variety identification.