基于共焦LIBS技术结合机器学习的矿石分类识别方法

doi:10.3964/j.issn.1000-0593(2023)03-0692-06

摘要
参考文献
相关文章 (15)

全文: PDF (17371 KB)
输出: BibTeX | EndNote (RIS)

摘要：矿物分类与识别是地质研究领域的重要内容，对地质勘探和环境演化的研究具有重要意义。然而，传统的矿石分类识别方法依靠专业人员通过矿石的外形及物理性质进行人工鉴定，主观性强，准确率低，激光诱导击穿光谱技术（LIBS）由于其元素“指纹”特性、灵敏度高以及快速在线检测的特点，非常适合用于地质研究领域。利用共焦激光诱导击穿光谱技术与机器学习结合，提高了矿石分类识别的精准度，利用共焦LIBS系统获得8种天然矿石样品（金矿、铜矿、银辉矿、赤铁矿、铝矿、方铅石、磷灰石以及闪锌矿）的光谱数据，采用主成分分析方法（PCA）对数据进行降维处理，并对降维后的数据采用线性判别分析（LDA）、最邻近规则（KNN）以及支持向量机(SVM)三种方法进行特征谱线的高精准分类识别。首先，采用标准铜片作为样品，对比了非共焦LIBS系统和共焦LIBS系统的稳定性及其对PCA主成分累计贡献率的影响，结果表明与非共焦LIBS系统相比，共焦LIBS系统的稳定性提升了63.75%，主成分累计贡献率提高了17.81%；然后，采用共焦LIBS系统获取上述8种矿石样品的光谱信息，并进行去噪等预处理，采用PCA对矿石特征数据进行提取，并保留累计贡献率达到99.4%的前10维特征空间；最后，将特征数据分别与LDA，KNN以及SVM结合构建分类模型，进行种类识别。结果表明，PCA方法与LDA和KNN方法结合的分类准确度分别为95.78%和92.58%，而与SVM相结合的方法，准确率可达到97.89%。因此，采用共焦激光诱导击穿光谱技术与PCA和SVM相结合的方法，可为地质勘探和矿物识别领域提供一种快速、高准确度的分类识别方式，具有广阔的应用前景。

关键词：激光诱导击穿光谱；共焦LIBS；主成分分析；机器学习

Abstract：Mineral classification and identification is an important area in the field of geological research, which is of great significance to geological exploration and environmental evolution. However, the traditional ore classification and identification methods rely on professionals to conduct manual identification through the shape and physical properties of the ore, which has strong subjectivity and low accuracy. Laser-induced breakdown spectroscopy (LIBS) is suitable for geological research due to itselement “fingerprint” characteristics, high sensitivity and fast on-line detection. In this paper, we use confocal laser-induced breakdown spectroscopy combined with machine learning to improve the accuracy of ore classification and recognition. The confocal LIBS system is used to obtain the spectral data of 8 natural ore samples (Gold, Copper, Silver, Hematite, Aluminum, Galena, Apatite and Sphalerite). Principal component analysis (PCA) is used to reduce the dimension of the data, Linear discriminant analysis (LDA), nearest neighbor rule (KNN) and support vector machine (SVM) are used for high-precision classification and recognition of feature spectral lines. Firstly, a standard copper is employed as the sample to conduct the comparison experiments between non confocal LIBS system and the confocal LIBS system for the stability and its influence on the cumulative contribution rate of PCA principal components. The results show that compared with non-confocal LIBS system, the stability of the confocal LIBS system is improved by 63.75%, and the cumulative contribution rate of principal components is increased by 17.81%. Then, the confocal LIBS system is used to obtain the spectral information of the above eight ore samples with data preprocessing, such as denoising. PCA is used to extract the ore feature data, and the first 10-dimensional feature space with a cumulative contribution rate of 99.4% is retained. Finally, the feature data are combined with LDA, KNN and SVM to build a classification model for classification and recognition. The experimental results show that the classification accuracy of PCA combined with LDA and KNN is 95.78% and 92.58% respectively, while that of SVM can reach 97.89%. Therefore, combining confocal laser-induced breakdown spectroscopy with PCA and SVM can provide a fast and accurate classification and recognition method for geological exploration and mineral recognition and has wide application prospects.

Key words：Laser-induced breakdown spectroscopy; Confocal LIBS; Principal component analysis; Machine learning

收稿日期: 2022-01-17 修订日期: 2022-04-21

中图分类号:

O433.1

基金资助: 国家自然科学基金仪器研制项目(61827826)，民用航天技术预先研究项目(D030207)资助

通讯作者: 崔晗 E-mail: han.cui@bit.edu.cn

作者简介: 苏云鹏，1997年生，北京理工大学光电学院硕士研究生 e-mail: whatuncle@163.com

引用本文:

苏云鹏，贺春景，李昂泽，徐可米，邱丽荣，崔晗. 基于共焦LIBS技术结合机器学习的矿石分类识别方法[J]. 光谱学与光谱分析, 2023, 43(03): 692-697.
SU Yun-peng, HE Chun-jing, LI Ang-ze, XU Ke-mi, QIU Li-rong, CUI Han. Ore Classification and Recognition Based on Confocal LIBS Combined With Machine Learning. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(03): 692-697.

链接本文:

https://www.gpxygpfx.com/CN/10.3964/j.issn.1000-0593(2023)03-0692-06 或 https://www.gpxygpfx.com/CN/Y2023/V43/I03/692

[1] Lanza N L, Wiens R C, Clegg S M, et al. Applied Optics, 2010, 49(13): 211.
[2] Chen T T, Zhang T L, Li H. Trends in Analytical Chemistry, 2020, 133(7): 116113.
[3] Trautner S, Lackner J, Spendelhofer W, et al. Analytical Chemistry, 2019, 91(8): 5200.
[4] Gaudiuso R, Ewusi-Annan E, Melikechi N, et al. Spectrochimica Acta Part B: Atomic Spectroscopy, 2018, 146(9): 106.
[5] Pagnotta S, Lezzerini M, Campanella B, et al. Spectrochimica Acta Part B: Atomic Spectroscopy, 2018, 146(20): 9.
[6] Teng G E, Wang Q Q, Kong J L, et al. Optics Express, 2019, 27(5): 6958.
[7] Nardecchia A, Fabre C, Cauzid J, et al. Analytica Chimica Acta, 2020, 1114(1): 66.
[8] YAN Meng-ge, DONG Xiao-zhou, LI Ying, et al（闫梦鸽，董晓舟，李颖，等）. Spectroscopy and Spectral Analysis（光谱学与光谱分析）, 2018, 38(6): 1874.
[9] YANG Yan-wei, ZHANG Li-li, HAO Xiao-jian, et al（杨彦伟，张丽丽，郝晓剑，等）. Infrared and Laser Engineering（红外与激光工程）, 2021, 50(5): 273.
[10] Alvarez J, Velasquez M, Myakalwar A K, et al. Journal of Analytical Atomic Spectrometry, 2019, 34(10): 2459.
[11] Tavares T R, Mouazen A M, Nunes L C, et al. Soil & Tillage Research, 2021, 216(12): 105250.
[12] Huang W H, He C J, Wang Y, et al. Journal of Analytical Atomic Spectrometry, 2020, 35(7): 2530.
[13] Wohlbrandt A, Hu N, Guerin S, et al. Computers and Fluids, 2016, 132(5): 46.
[14] Trisnawati N L P, Krisandi A, Widagda I G A, et al. Journal of Physics: Conference Series, 2020, 1572(8): 012002.
[15] Guo Y M, Tang Y, Du Y, et al. Plasma Science and Technology, 2018, 20(6): 2058.
[16] Gyftokostas N, Nanou E, Stefas D, et al. Molecules, 2021, 26(5): 1241.
[17] Xu Y L, Meng L W, Chen X J, et al. Plasma Science and Technology, 2021, 23(8): 133.
[18] LI Ang-ze, WANG Xian-shuang, XU Xiang-jun, et al（李昂泽，王宪双，徐向君，等）. Chinese Optics（中国光学）, 2019, 12(5): 1139.