微塑料
拉曼光谱
合并(版本控制)
主成分分析
加权
算法
人工智能
高光谱成像
计算机科学
模式识别(心理学)
生物系统
化学
光学
声学
物理
环境化学
生物
情报检索
作者
Yunlong Luo,Christopher T. Gibson,Clarence Chuah,Youhong Tang,Ravi Naidu,Cheng Fang
标识
DOI:10.1016/j.scitotenv.2022.158293
摘要
The characterisation of microplastics is still difficult, and the challenge is even greater for nanoplastics. A possible source of these particles is the scratched surface of a non-stick cooking pot that is mainly coated with Teflon. Herein we employ Raman imaging to scan the surfaces of different non-stick pots and collect spectra as spectrum matrices, akin to a hyperspectral imaging process. We adjust and optimise different algorithms and create a new hybrid algorithm to extract the extremely weak signal of Teflon microplastics and particularly nanoplastics. We use multiple characteristic peaks of Teflon to create several images, and merge them to one, using a logic-based algorithm (i), in order to cross-check them and to increase the signal-noise ratio. To differentiate the varied peak heights towards image merging, an algebra-based algorithm (ii) is developed to process different images with weighting factors. To map the images via the whole set of the spectrum (not just from the individual characteristic peaks), a principal component analysis (PCA)-based algorithm (iii) is employed to orthogonally decode the spectrum matrix to the PCA spectrum and PCA intensity image. To effectively extract the Teflon spectrum information, a new hybrid algorithm is developed to justify the PCA spectra and merge the PCA intensity images with the algebra-based algorithm (PCA/algebra-based algorithm) (iv). Based on these developments and with the help of SEM, we estimate that thousands to millions of Teflon microplastics and nanoplastics might be released during a mimic cooking process. Overall, it is recommended that Raman imaging, along with the signal recognition algorithms, be combined with SEM to characterise and quantify microplastics and nanoplastics.
科研通智能强力驱动
Strongly Powered by AbleSci AI