摘要
针对目前存在的技术未达到既快速又准确地将食用油种类分类识别出来,为此提出一种基于激光诱导荧光技术并结合间隔偏最小二乘法(iPLS)进行食用油的快速识别方法。实验选择常见4种品牌的5种油,每种120个样本,共600个。首先采用激光诱导荧光系统采集荧光光谱,然后通过间隔偏最小二乘法(iPLS)算法筛选出特征波段,随后采用随机划分法划分训练集和测试集作为BP神经网络的输入进行建模,通过比较发现划分为12个子区间时所筛选出的344个波长点的预测准确率达到了100%,较全波段的91.68%准确率有明显地提升。实验结果表明,运用i PLS波段筛选得到的特征波段再结合BP创建的模型可以实现食用油的快速检测分类,具有良好的市场应用前景。
In view of the fact that the existing technologies fail to achieve rapid and accurate classification of edible oils,a rapid identification method for edible oils based on laser-induced fluorescence technology combined with interval partial least squares(iPLS)is proposed.Five oils of four common brands are selected in the experiment,each of them contains 120 samples and the total of 600 oils.Firstly,the laser-induced fluorescence system is used to collect the fluorescence spectrum.Then,the feature band is selected by interval partial least squares(iPLS)algorithm.The training set and the test set are divided by random partition method as the input of BP neural network for modeling.The prediction accuracy of the 344 wavelength points screened among the divided 12 sub-intervals reaches 100%,which is a significant improvement over the 91.68%accuracy of the full band.The experimental results show that using the characteristic bands obtained by i PLS band screening combined with the model created by BP can realize rapid detection and classification of edible oil,which has a good market application prospect.
作者
周孟然
孙磊
卞凯
胡锋
来文豪
余道洋
闫鹏程
ZHOU Mengran;SUN Lei;BIAN Kai;HU Feng;LAI Wenhao;YU Daoyang;YAN Pengcheng(School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan Anhui 232000,China;College of Mechanics and Optoeletronic Physics,Anhui University of Science and Technology,Huainan Anhui 232000,China)
出处
《激光杂志》
北大核心
2020年第7期13-17,共5页
Laser Journal
基金
国家重点研发计划(No.2018YFC0604503)
安徽省自然科学基金青年项目(No.1808085QE157)
国家安全生产重大事故防治关键技术科技项目(No.anhui-0001-2016AQ)。
