摘要
近年来由于塑料的大量使用和排放,这些塑料经环境作用破碎变成微塑料大量汇聚到海洋中,导致海洋中聚集大量微塑料。微塑料形状较小,难以识别其来源与种类。激光拉曼探测技术具有快速、无损、且各物质指纹峰明显易被精确识别等优点,近年来被广泛应用。本文基于拉曼光谱探测技术,提出了一种结合小波处理、随机森林算法实现海水中微塑料快速识别的智能分类方法。针对六种典型的海水微塑料标准样品(丙烯腈(A)-丁二烯(B)-苯乙烯(S)的三元共聚物(ABS)、聚酰胺(PA)、聚对苯二甲酸乙二醇酯(PET)、聚丙烯(PP)、聚苯乙烯(PS)、聚氯乙烯(PVC)),采用激光拉曼探测技术进行光谱数据收集,对获取的拉曼光谱采用小波基为DB7、分解次数为3的小波,标准差归一化进行了拉曼光谱预处理。为了提高识别速度,同时还需要对光谱数据进行数据压缩预处理,分别进行了数据压缩点为64,128,256,512和1024点的数据压缩比较,它们的决策树算法识别精度分别为91.51%,91.67%,92.35%,93.17%和93.21%,随机森林算法识别精度分别为93.12%,93.92%,94.83%,96.81%和96.81%,实验结果表明,微塑料的拉曼光谱压缩为512点时为效率和精度的最佳压缩点,可以为实际工程应用中微塑料拉曼数据压缩提供参考。分别采用决策树、随机森林两种算法进行微塑料拉曼光谱识别研究。研究结果表明,基于拉曼光谱数据,随机森林算法的识别微塑料交叉验证精度高于决策树算法。为进一步提高识别精度,进行了模型参数(折次k)优化研究,采用经过优化后的模型参数(k=20),随机森林算法识别微塑料的交叉验证精度可以达到97.24%。可以为实际海水中微塑料的快速识别提供技术参考。
Due to a large amount of use and discharge of plastics,these plastics are broken into microplastics by the environmental effect and gather in the ocean in large quantities,leading to the accumulation of a large number of microplastics in the ocean,inrecent year.Microplastics are small in shape and difficult to identify their source and type.Laser Raman detection technology has been widely used in recent years which have fast,nondestructive and easy identification.In this paper,based on Raman spectral detection technology,an intelligent classification method combining wavelet processing and random forest algorithm is proposed to realize the rapid recognition of microplastics in seawater.The spectral data were collected by using laser Raman detection technology from six typical seawater microplastics standard samples(ABS,PA,PET,PP,PS,PVC),and the obtained spectra were pretreated by wavelet base DB7 and decomposition times 3 and standard deviation normalization.In order to improve the recognition speed,the spectral data is compressed at the same time.The data are respectively compressed to 64,128,256,512 and 1024 points,and their decision tree algorithm identification accuracy was 91.51%,91.67%,92.35%,93.17%and 93.21%respectively.The random forest algorithm identification accuracy was 93.12%,93.92%,94.83%,96.81%and 96.81%,respectively.The experimental results show that the Raman spectral compression of microplastics is the best compression point for efficiency and precision when the Raman spectral compression is 512 points,which can provide a reference for the Raman data compression of microplastics in practical engineering applications.Two recognition algorithms,decision tree and random forest,were used to study the Raman spectrum recognition of microplastics.The results show that the cross-validation accuracy of the random forest is higher than that of the decision tree.In order to further improve the identification accuracy,the model parameter optimization was carried out,and the cross-validation accuracy of the
作者
杨思节
冯巍巍
蔡宗岐
王清
YANG Si-jie;FENG Wei-wei;CAI Zong-qi;WANG Qing(Harbin Institute of Technology(Weihai),Weihai 264200,China;CAS Key Laboratory of Coastal Environmental Processes and Ecological Remediation,Yantai Institute of Coastal Zone Research,Chinese Academy of Sciences,Yantai 264003,China;Center for Ocean Mega-Science,Chinese Academy of Sciences,Qingdao 266071,China;University of Chinese Academy of Sciences,Beijing 100049,China)
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2021年第8期2469-2473,共5页
Spectroscopy and Spectral Analysis
基金
国家重点研发计划项目(2019YFD0901101)
山东省重点研发计划项目(2019JZZY010810)资助。
关键词
微塑料
激光拉曼
小波分析
决策树
随机森林
Microplastics
Laser Raman
Wavelet analysis
Decision tree
Random forest
