基于特征光谱的草莓品种快速鉴别

Discrimination of Strawberries Varieties Based on Characteristic Spectrum

摘要:

应用近红外光谱技术对草莓品种快速鉴别进行了研究。通过比较7种预处理方法，确定SNV+MAF+2D组合预处理方法最佳。采用相关系数阈值法提取了14个与草莓品种信息相关的特征光谱。建立了基于548～562nm范围内14个特征光谱的PLS—ANN、PLSR、PCR 3种校正模型。经预测集样品验证，主成分数为9时的PLS—ANN模型最优，其预测相关系数和预测均方根误差分别为0.9677和0.461。结果表明，通过提取少量特征光谱建立PLS—ANN校正模型能实现草莓品种的快速鉴别。

Abstract:

Near infrared spectroscopy (NIRS) technology was applied to discriminate the variety of strawberries rapidly. Compared with seven kinds of preprocessing methods, “SNV+MAF+2D” was ascertained as the optimal scheme. Totally 14 characteristic spectrums related to strawberries information were extracted by the correlation coefficient and threshold value method. Three kinds of correction models of PLS—ANN, PLSR and PCR based on wavelength ranging from 548nm to 562nm were established. Verified by the prediction set samples, the optimal correction model was PLS—ANN with nine principal components. Prediction correlation coefficientand root mean square error of prediction(RMSEP) of the PLS—ANN correction model was 0.9677 and 0.461. The results showed that a few characteristic spectrum extracted to establish PLS—ANN correction model would achieve rapid discrimination of strawberries.