基于LIBS技术结合PCA-SVM机器学习对猪肉部位的识别研究  

Identification of Pork Parts Based on LIBS Technology Combined With PCA-SVM Machine Learning

作　　者：许毓婷 孙浩然 高勋[1] 郭凯敏[3] 林景全[1] XU Yu-ting;SUN Hao-ran;GAO Xun;GUO Kai-min;LIN Jing-quan(College of Science,Changchun University of Science and Technology,Changchun 130022,China;College of Mechanical and Electrical Engineering,Changchun University of Technology,Changchun 130012,China;College of Physical Science and Technology,Baotou Normal University,Baotou 014030,China)

机构地区：[1]长春理工大学理学院,吉林长春130022 [2]长春工业大学机电工程学院,吉林长春130012 [3]包头师范学院物理科学与技术学院,内蒙古包头014030

出　　处：《光谱学与光谱分析》2021年第11期3572-3576,共5页Spectroscopy and Spectral Analysis

基　　金：国家自然科学基金项目(61575030);吉林省科技厅项目(20180201046YY,20180101283JC,20200301042RQ);吉林省教育厅项目(JJKH20190539KJ)资助。

摘　　要：近年来,激光诱导击穿光谱(LIBS)与算法相结合分类、识别生物组织逐渐兴起。由于猪肉各部位组织光谱特性相似,仅通过分析光谱信息很难达到准确识别的效果,采用来自于同一个体、四个不同部位的猪肉进行研究,并将其进行切片、压平,应用LIBS技术对4种部位的组织(里脊,梅花,前腿,五花)进行了样品光谱的采集,每种样品采集100幅光谱进行分析,选取Ca, Na, K等6条谱线进行了初步光谱分析,观测谱线发现除脂肪含量较多的五花组织C—N以及C含量较其他组织高以外,其他组织很难区分,进一步对这6个成分进行主成分分析(PCA),得到PC1, PC2, PC3累计贡献率达到95%,通过特征分数作为支持向量机(SVM)模型输入源,建立SVM分类模型,得到几种部位样品的混淆矩阵图,通过观察混淆矩阵可以清楚分辨出每个种类样品的分类整准确率,发现四种样品准确率分别为96%, 98%, 97%, 100%,平均准确率达到了97%以上。研究证明LIBS结合PCA-SVM可作为一种快速鉴别猪肉不同组织部位的检测方法。In recent years, laser-induced breakdown spectroscopy(LIBS) is gradually emerging to classify and identify biological organizations by combining them with algorithms. Due to each part of the pork similar spectral characteristics, it is difficult to achieve accurate identifications only through the analysis of the effect of spectral information, so in this paper studied pork from four different parts of the same individual and sliced and planished them, then applied LIBS technology on four parts, i.e., the organization, fillet, plum flower, and front legs. 100 specimens of each sample were collected and the spectrum analysis was conducted. A preliminary analysis of the spectrum was performed on Ca, Na, K and 6 lines. It was found that other tissues were difficult to distinguish except for the C—N tissue of plum flower with more fat content and higher C content than other tissues, so the Principal Component Analysis(PCA) on these 6 principal components was carried out. The cumulative contribution rate of PC1, PC2 and PC3 reached 95%. The Support Vector Machines(SVM) classification model was established by employing feature scores as the input source of SVM model, and the confusion matrix diagram of these samples got obtained. Through observation of the confusion matrix, the classification accuracy of each type of samples could be clearly distinguished. The results showed that the accuracy of the four samples was 96%, 98%, 97% and 100%, respectively, with an average accuracy surpassing 97%. The study proved that LIBS combined with PCA-SVM can be used as a fast identification method for different parts of pork tissues.

关 键 词：激光诱导击穿光谱 主成分分析 支持向量机 猪肉组织 

分 类 号：O433.4[机械工程—光学工程]