EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Lightweight convolutional neural network for nicotine prediction in tobacco by near-infrared spectroscopy

Di Wang, Fengyuan Zhao, Rui Wang, Junwei Guo, Cihai Zhang, Huimin Liu, Yongsheng Wang, Guohao Zong, Le Zhao, Weihua Feng

Frontiers in Plant Science · 12.05.2023 · 14

DOI: 10.3389/fpls.2023.1138693 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Plant Science, online erschienen am 12.05.2023. Verfasst von Di Wang und 9 weiteren Autor:innen. Beteiligte Einrichtung: Zhengzhou Tobacco Research Institute of China National Tobacco Corporation (CNTC), Zhengzhou, China. Die Arbeit wurde bisher 20-mal zitiert (Crossref) und nennt 43 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The content of nicotine, a critical component of tobacco, significantly influences the quality of tobacco leaves. Near-infrared (NIR) spectroscopy is a widely used technique for rapid, non-destructive, and environmentally friendly analysis of nicotine levels in tobacco. In this paper, we propose a novel regression model, Lightweight one-dimensional convolutional neural network (1D-CNN), for predicting nicotine content in tobacco leaves using one-dimensional (1D) NIR spectral data and a deep learning approach with convolutional neural network (CNN). This study employed Savitzky–Golay (SG) smoothing to preprocess NIR spectra and randomly generate representative training and test datasets. Batch normalization was used in network regularization to reduce overfitting and improve the generalization performance of the Lightweight 1D-CNN model under a limited training dataset. The network structure of this CNN model consists of four convolutional layers to extract high-level features from the input data. The output of these layers is then fed into a fully connected layer, which uses a linear activation function to output the predicted numerical value of nicotine. After the comparison of the performance of multiple regression models, including support vector regression (SVR), partial least squares regression (PLSR), 1D-CNN, and Lightweight 1D-CNN, under the preprocessing method of SG smoothing, we found that the Lightweight 1D-CNN regression model with batch normalization achieved root mean square error (RMSE) of 0.14, coefficient of determination ( R 2 ) of 0.95, and residual prediction deviation (RPD) of 5.09. These results demonstrate that the Lightweight 1D-CNN model is objective and robust and outperforms existing methods in terms of accuracy, which has the potential to significantly improve quality control processes in the tobacco industry by accurately and rapidly analyzing the nicotine content.

Schlagwörter

Stichwörter der Autor:innen

convolutional neural networklightweightnear-infrared spectroscopynicotinetobacco

Autor:innen und Institutionen

  • Zhengzhou Tobacco Research Institute of China National Tobacco Corporation (CNTC), Zhengzhou, China.
  • School of Mathematics and Statistics, Zhengzhou University, Zhengzhou, China.
  • Technology Center of China Tobacco Guizhou Industrial Co., Ltd., Guiyang, China.

Zitieren

APA

Wang, D., Zhao, F., Wang, R., Guo, J., Zhang, C., Liu, H., Wang, Y., Zong, G., Zhao, L., & Feng, W. (2023). A Lightweight convolutional neural network for nicotine prediction in tobacco by near-infrared spectroscopy. Frontiers in Plant Science, 14. https://doi.org/10.3389/fpls.2023.1138693

Vancouver

Wang D, Zhao F, Wang R, Guo J, Zhang C, Liu H, et al. A Lightweight convolutional neural network for nicotine prediction in tobacco by near-infrared spectroscopy. Frontiers in Plant Science. 2023;14. doi:10.3389/fpls.2023.1138693

BibTeX
@article{Wang2023,
  author = {Wang, Di and Zhao, Fengyuan and Wang, Rui and Guo, Junwei and Zhang, Cihai and Liu, Huimin and Wang, Yongsheng and Zong, Guohao and Zhao, Le and Feng, Weihua},
  title = {A Lightweight convolutional neural network for nicotine prediction in tobacco by near-infrared spectroscopy},
  journal = {Frontiers in Plant Science},
  year = {2023},
  volume = {14},
  doi = {10.3389/fpls.2023.1138693}
}
Kostenlos starten →