| 接昭玮1, 李绅2, 汪睿璇3，王继芬1，张震1, 徐晓杰4，周娣4，石学军4.红外光谱快速识别食用植物油种类的研究[J].中国油脂,2023,48(7):.[JIE Zhaowei1, LI Shen2, WANG Ruixuan3, WANG Jifen1, ZHANG Zhen1,
XU Xiaojie4, ZHOU Di4, SHI Xuejun4.Fast identification of edible vegetable oil kinds by infrared spectroscopy[J].China Oils and Fats,2023,48(7):.] |
| 红外光谱快速识别食用植物油种类的研究 |
| Fast identification of edible vegetable oil kinds by infrared spectroscopy |
| |
| DOI： |
| 中文关键词: 食用植物油 红外光谱 特征提取 CARS-CS-ELM 随机森林 分类识别 |
| 英文关键词:vegetable oil infrared spectroscopy feature extraction CARS-CS-ELM random forest classification and recognition |
| 基金项目:中央高校基本科研业务费专项资金资助（2021JKF208) |
|
| 摘要点击次数: 1354 |
| 全文下载次数: 1401 |
| 中文摘要: |
| 为实现食用植物油种类的快速无损识别，为公安实战中打击“食药环”犯罪提供参考，借助衰减全反射-傅里叶变换红外光谱技术对不同类别、品牌食用植物油进行了多层次分类识别工作。采用标准正态变换（SNV）和一阶导数预处理消除基线和其他背景干扰，使得重叠峰发生分离，从而提高检测的分辨率和灵敏度，利用竞争性自适应重加权算法（CARS）提取特征波长，结合基于布谷鸟搜索算法优化的极限学习机（CS-ELM）模型对不同种类和品牌的食用植物油进行分类识别，同时对比随机森林模型与CARS-CS-ELM融合模型在食用植物油快速分类检测方面的准确率。结果表明，基于CARS-CS-ELM融合模型对3类植物油样本总体进行分类，其分类准确率达到8519%，其中小磨香油、花生油、玉米油样本训练集的品牌分类准确率依次为92.5%、100%、967%，测试集品牌分类准确率均为100%，而随机森林模型的9个品牌食用植物油分类准确率仅为80%。综上，CARS-CS-ELM融合模型对食用植物油快速分类识别效果较好，可为食用植物油的无损快速检验提供一定的参考与借鉴。 |
| 英文摘要: |
| In order to achieve rapid non-destructive identification of edible vegetable oil kinds and provide reference for the fight against food, drug and environment crimes in public security operations, the attenuated total reflection-Fourier transform infrared spectroscopy analysis technology was used to carry out multi-level classification and recognition of different kinds and brands of edible vegetable oils. The experiment used standard normal variation (SNV) and first-order derivative preprocessing to eliminate baseline and other background interference, resulting in the separation of overlapping peaks and improving detection resolution and sensitivity. The competitive adaptive reweighted sampling(CARS) algorithm was used to extract feature wavelengths, combining the extreme learning machine (CS-ELM) model optimized based on cuckoo search algorithm to classify and identify different kinds and brands of edible vegetable oils. The accuracy of random forest model and CARS-CS-ELM model were compared in the rapid classification and detection of edible vegetable oils. The results showed that the overall classification accuracy of three types of vegetable oil samples based on CARS-CS-ELM model reached 85.19%, among which the brand classification accuracy of the training set of sesame oil, peanut oil and corn oil was 92.5%, 100% and 96.7% respectively, and that of the test set was 100%, while the classification accuracy of nine brands of edible vegetable oil in random forest model was only 80%. In summary, the CARS-CS-ELM model has good performance in rapid classification and recognition of edible vegetable oils, and can provide certain reference for non-destructive and rapid testing of edible vegetable oils. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
