中文摘要：

提出了一种基于可见-近红外光谱技术快速、无损鉴别杂交稻种品系与真伪的新方法。采集了5种稻种的光谱数据，各获取32个样本，随机分成训练集（125份）和检验集（35份）。光谱经S．Golay平滑和标准归一化（SNV）处理后，以主成分分析法（PCA）降维。将降维所得的前9个主成分作为新变量。分别用模糊模式识别、BP-神经网络、Fisher多类线性判别以及Bayes多类逐步判别四种方法进行分析。对35个未知样的预测结果说明可见-近红外技术进行杂交稻种品系与真伪的快速、无损鉴别是可行的，且PCA结合Bayes多类逐步判别是一种优选方法。

英文摘要：

A fast and non-invasive method based on visible-near infrared reflectance spectroscopy was put forward for discriminating lines and authenticity of hybrid rice seed. Five different varieties of rice seed were analyzed using a FieldSpec 3 visible-near infrared spectrometer,and 32 samples were used for each variety of rice. All samples were divided randomly into two groups, one group with 125 samples used as calibrated set, and another with 35 samples used as validated set. The samples data were pretreated with the methods of S. Golay smoothing and standard normal variable （SNV）,and then analyzed by principal component analysis （PCA）. The anterior 9 principal components computed by PCA were used as the new variables, and analysised by Fuzzy pattern recognition, back-propagation artifi- cial neural network （ANN-BP） ,Fisher multi-types linear discriminant and Bayes multi-types stepwise discriminant. Then, the 35 unknown samples in the validated set were predicted. Therefore, the feasibility of discriminating the lines and authenticity of hybrid rice seed in rapid and non-invasive way by visible-near spectra technology was proved, and PCA combined with Bayes multi-types stepwise discriminant was confirmed as a preference method.