A novel way to rapidly monitor microplastics in soil by hyperspectral imaging technology and chemometrics
A novel way to rapidly monitor microplastics in soil by hyperspectral imaging technology and chemometrics
复制标题
利用高光谱成像技术和化学计量学快速监测土壤中微塑料的新方法
DOI：
10.1016/j.envpol.2018.03.026
复制
发表时间：
2018
影响因子：
8.9
通讯作者：
Wu Fengchang
中科院分区：
文献类型：
--
作者：
Shan Jiajia;Zhao Junbo;Liu Lifen;Zhang Yituo;Wang Xue;Wu Fengchang
本研究探讨了高光谱成像技术作为一种直接有效检测土壤中微塑料污染的可能方法。从含有微塑料、鲜叶、枯叶、岩石和干枝等不同物质的土壤样品中获得波长在400 ~ 1000 nm之间的高光谱图像。利用支持向量机（SVM）、马氏距离（MD）和最大似然（ML）算法等监督分类算法，从高光谱图像中识别微塑料。为了研究粒径和颜色的影响，从土壤中提取了粒径范围为1-5 mm和0.5-1 mm的白色聚乙烯和黑色聚乙烯颗粒。结果表明，SVM是最适用于土壤中白色PE的检测方法，在1 ~ 5 mm和0.5 ~ 1 mm粒径范围内PE的检测精度分别为84%和77%。对于1 ~ 5 mm和0.5 ~ 1 mm的颗粒，SVM的黑色PE检测精度分别为58%和76%。以饮料瓶、瓶盖、橡胶、包装袋、衣架、塑料夹等6种家用聚合物为例进行验证，对1 ~ 5 mm和0.5 ~ 1 mm的塑料微粒，聚合物的分类精度分别为79% ~ 100%和86% ~ 99%。结果表明，高光谱成像技术是直接测定和可视化土壤表面0.5 ~ 5 mm粒径微塑料的一种有潜力的技术。
Hyperspectral imaging technology has been investigated as a possible way to detect microplastics contamination in soil directly and efficiently in this study. Hyperspectral images with wavelength range between 400 and 1000 nm were obtained from soil samples containing different materials including microplastics, fresh leaves, wilted leaves, rocks and dry branches. Supervised classification algorithms such as support vector machine (SVM), mahalanobis distance (MD) and maximum likelihood (ML) algorithms were used to identify microplastics from the other materials in hyperspectral images. To investigate the effect of particle size and color, white polyethylene (PE) and black PE particles extracted from soil with two different particle size ranges (1–5 mm and 0.5–1 mm) were studied in this work. The results showed that SVM was the most applicable method for detecting white PE in soil, with the precision of 84% and 77% for PE particles in size ranges of 1–5 mm and 0.5–1 mm respectively. The precision of black PE detection achieved by SVM were 58% and 76% for particles of 1–5 mm and 0.5–1 mm respectively. Six kinds of household polymers including drink bottle, bottle cap, rubber, packing bag, clothes hanger and plastic clip were used to validate the developed method, and the classification precision of polymers were obtained from 79% to 100% and 86%–99% for microplastics particle 1–5 mm and 0.5–1 mm respectively. The results indicate that hyperspectral imaging technology is a potential technique to determine and visualize the microplastics with particle size from 0.5 to 5 mm on soil surface directly.