摘要
为了验证激光诱导击穿光谱技术对不同陈化年份的样品进行快速分类鉴别的可行性,采用激光诱导击穿光谱技术和随机森林相结合研究样品分类问题。采用波长为1064 nm的纳秒脉冲激光器作为激发光源,能量选取50 mJ,通过光谱仪进行光谱采集,选用样品中常量元素K、Ca、Mg、Fe、Na、Cu的光谱数据,将上述元素谱线的峰值强度进行归一化处理后作为特征变量,通过随机森林建立分类模型,划分训练集与测试集进行训练,分类准确率的算术平均值达到98.89%。研究结果表明,激光诱导击穿光谱技术结合随机森林算法可降低检测灵敏度低和噪声干扰带来的影响,提高分类准确性和可靠性。
In order to verify the feasibility of laser-induced breakdown spectroscopy for rapid classifi⁃cation and identification of samples with different ageing years,a combination of laser-induced breakdown spectroscopy and random forest was used to study the sample classification problem.A nanosecond pulsed laser with a wavelength of 1064 nm was used as the excitation light source with an energy of 50 mJ,and the spectra were collected by a spectrometer,and the spectral data of the macronutrient elements K,Ca,Mg,Fe,Na,and Cu in the samples were selected,and the peak intensities of the spectral lines of the above ele⁃ments were normalised to serve as the feature variables,and the classification model was established by the random forest,and the training and testing sets were divided for The arithmetic mean of classification accu⁃racy reaches 98.89%.The results show that laser-induced breakdown spectroscopy technique combined with random forest algorithm can reduce the effects of low detection sensitivity and noise interference,and improve classification accuracy and reliability.
作者
陶翰飞
谭嘉城
赵静怡
李业秋
岱钦
TAO Hanfei;TAN Jiacheng;ZHAO Jingyi;LI Yeqiu;DAI Qin(College of Science,Shenyang Ligong University,Shenyang 110159,China)
出处
《微处理机》
2025年第4期8-12,共5页
Microprocessors
基金
辽宁省教育厅高等学校基本科研项目(LJKZ0262)。
关键词
激光诱导击穿光谱
年份鉴别
快速分类
机器学习
laser-induced breakdown spectroscopy
year identification
rapid classification
machine learning
