运用激光诱导击穿光谱(LIBS)技术对赣南脐橙橙汁进行了快速绿色鉴别。实验分别测定了健康和黄龙病脐橙果汁的糖度及Ca、K、Zn元素含量,并分析了糖度及元素含量差异。采集了脐橙果汁的LIBS光谱数据,运用九点平滑(9SM)法并结合多元散射校正(MSC)对数据进行了预处理,最后运用主成分分析(PCA)法并结合多层感知器(MLP)神经网络和径向基函数(RBF)神经网络模型对健康和黄龙病脐橙进行了快速判别。结果表明,PCA-MLP模型对健康和黄龙病脐橙的判别效果优于PCA-RBF模型,其训练集对健康脐橙和黄龙病脐橙的判别准确率分别为93.8%和93.4%,预测集对健康脐橙和黄龙病脐橙的判别准确率分别为93.9%和94.8%。LIBS检测结果证明了黄龙病导致脐橙果肉品质发生了变化;进一步利用光谱预处理方法和分类模型,从品质上区分了黄龙病脐橙果汁和健康脐橙果汁,提高了出厂橙汁的产品合格率。

The laser induced breakdown spectroscopy (LIBS) method is used for the rapid and green identification of Gannan navel orange juices. The sugar contents and Ca,K,Zn element contents of healthy and Huanglongbing navel oranges are experimentally measured. In addition, the differences in sugar and element contents are analyzed. The LIBS data of navel orange juice is first collected, which is then preprocessed by the nine-point smoothing (9SM) method combined with multivariate scattering correction (MSC). Finally, the principal component analysis (PCA) method combined with the multi-layer perceptron (MLP) neural network and radial basis function (RBF) neural network model is used for rapid identification of healthy and Huanglongbing navel orange juice. The results show that the PCA-MLP model is superior to the PCA-RBF model in the identification effect of healthy and Huanglongbing navel oranges. The identification accuracies of healthy and Huanglongbing navel oranges on the training dataset are 93.8% and 93.4%, respectively. In contrast, the identification accuracies of healthy navel oranges and Huanglongbing navel oranges on the prediction dataset are 93.9% and 94.8%, respectively. The LIBS detection results confirm that the Huanglongbing results in the change in pulp quality of navel oranges. The further spectral preprocessing and the classification model are used to distinguish the juices of Huanglongbing oranges and healthy navel oranges in quality and thus the product qualification ratio of factory orange juices is increased.