激光与光电子学进展, 2020, 57 (20): 203001, 网络出版: 2020-10-17   

核极限学习机和激光诱导荧光技术在食用油识别中的应用 下载: 978次

Application of Kernel Extreme Learning Machine and Laser Induction Fluorescence Technique in Edible Oil Identification
作者单位
安徽理工大学电气与信息工程学院, 安徽 淮南 232001
摘要
针对现有的食用油检测技术无法快速、准确地识别市售食用油的问题,提出了一种快速辨识食用油的方法。采用激光诱导荧光技术(LIF)获取油样的荧光光谱数据,然后采用主成分分析法提取光谱数据的特征信息,之后采用飞蛾-扑火优化器和核极限学习机相结合的算法建立多元分类学习模型,最后用该模型识别油样的类别。实验油样选取5种样本,每种样本采集150组荧光光谱,然后随机抽取600个样本用于学习模型的训练,剩余的150个用于测试训练好的模型。结果表明:在测试集上的平均分类准确率方面,该模型与极限学习机、反向传播神经网络相差不大,但该模型分类准确率的标准差远小于其他两种模型。这说明所建模型具有较稳定的分类性能,可以满足快速鉴别食用油的要求。
Abstract
Existing edible oil detection technology cannot quickly and accurately identify edible oils sold in markets. Hence, in this paper, we propose a quick method of identifying edible oils. Fluorescence spectrum data of oil samples were obtained using the laser induction fluorescence(LIF) technique. Principal component analysis was used to extract characteristic information. Next, a multiclassification learning model was developed through the fusion algorithm of moth-flame optimization and kernel extreme learning machine (KELM) to identify the type of oil samples. Five types of oil samples were selected for experimental purposes, and 150 groups of fluorescence spectra were collected from each sample. Next, 600 samples were randomly selected to train the learning model, and the remaining 150 samples were used to test the trained model. Experimental results show that KELM model , extreme learning machine model and back propagation neural network model have similar average classification accuracy on the test set. However, the standard deviation of KELM model is less than those of other two models. This shows that KELM model has a stable classification performance and can quickly identify edible oils.

周孟然, 王锦国, 宋红萍, 胡锋, 来文豪, 卞凯. 核极限学习机和激光诱导荧光技术在食用油识别中的应用[J]. 激光与光电子学进展, 2020, 57(20): 203001. Mengran Zhou, Jinguo Wang, Hongping Song, Feng Hu, Wenhao Lai, Kai Bian. Application of Kernel Extreme Learning Machine and Laser Induction Fluorescence Technique in Edible Oil Identification[J]. Laser & Optoelectronics Progress, 2020, 57(20): 203001.

本文已被 2 篇论文引用
被引统计数据来源于中国光学期刊网
引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!