用傅里叶变换红外光谱（FTIR）和逐步判别分析法对油菜籽的品种和产地进行鉴别研究。 测试了来自5个产地17个油菜籽品种188份油菜籽皮样品的红外光谱, 按产地和品种选择训练样本, 每个品种选择5份为训练样本, 剩下的作为测试样本, 选择1　800～950 cm-1范围的光谱信息, 在SPSS20.0的判别分析模块中采用逐步法, Fisher线性判别准则, 分别建立油菜籽品种和产地的鉴别模型来判别样品的品种和产地； 比较了逐步判别分析中五种筛选建模信息的算法所建立的模型对实验样品品种和产地的鉴别效果。 基于油菜籽皮红外光谱信息的品种判别分析, 五种筛选变量的算法所建立的模型都能较好的识别油菜籽的品种, “Wilks’ Lambda”法所建模型的识别效果最好, 判别正确率为97.9%。 基于油菜籽皮红外光谱信息的产地判别分析, 五种筛选变量的算法所建立的产地判别模型也都能较好的识别油菜籽的产地, “Unexplained variance”法所建模型的识别效果最好, 判别正确率为98.4%。 研究结果表明, 基于油菜籽皮中红外光谱信息的品种和产地判别分析, 有望成为油菜籽品种和产地鉴别的方法。

The planting area and varieties of rapeseeds were studied based on Fourier transform infrared spectroscopy (FTIR) combining with stepwise discriminatory analysis. The FTIR of 188 rapeseed skin samples of 17 varieties from 5 planting areas was obtained and processed by OMINC 8.0 software. In order to distinguish their planting area and variety, 5 spectra in the range of 1 800～950 cm-1 for each variety from each producing area were selected as training samples arbitrarily with Fisher linear discriminatory criterion which was used to build model by means of stepwise discriminatory analysis by SPSS20.0 software. All the five discriminatory models based on Wilks’Lambda, Unexplained variance, Mahalanobis distance, Smallest F value and Rao’s v can distinguish variety and planting area well, and the best one for identifying variety was on the base of Wilks’Lambda, which yielded correct rate of 97.9%, and the best one for distinguishing planting area was on the base of Unexplained variance with a correct rate of 98.4%. The results showed that discriminatory analysis based on the infrared spectrum of rapeseed skin is an efficient method for identification of rapeseed variety and planting area.