基于多维度傅里叶红外光谱与两种神经网络模型对昭通苹果的鉴别分析

doi:10.3964/j.issn.1000-0593(2025)06-1543-08

Abstract
References
Related Citation (15)

Download: PDF (12249 KB)
Export: BibTeX | EndNote (RIS)

Abstract The article presents an analysis of eight different varieties of Zhaotong apples using Fourier Transform Infrared Spectroscopy (FTIR) and Two-Dimensional Correlation Infrared Spectroscopy (2D-IR). Additionally, Convolutional Neural Networks (CNN) and Radial Basis Function (RBF) neural networks were utilized for their identification. The Fourier Transform Infrared Spectroscopy (FTIR) analysis of eight Zhaotong apple varieties revealed prominent absorption peaks within the spectral ranges of 3 500～2 850, 1 650～1 400, and 1 250～800 cm^-1, signifying their abundance in sugars, vitamins, amino acids, lipids, organic acids, phenols, flavonoids, and various other compounds. However, a notable similarity was observed across the spectra of these apples, with only subtle variations in peak intensities and positions. Consequently, relying solely on spectral characteristics to differentiate between these eight varieties is impractical. Using temperature as a perturbation, we collected dynamic spectra of eight distinct apple varieties and subjected them to 2D-IR analysis within the spectral range of 800～1 800 cm^-1. The synchronous spectrum derived from this analysis underscores that as temperature escalates, distinct autopeaks emerge in proximity to 1 010 and 1 642 cm^-1. These peaks serve as indicators of varying degrees of decomposition occurring within esters, acids, and proteins present in apples, with esters and acids undergoing more pronounced decomposition compared to proteins. Notably, among these eight samples, Red Fuji Apples exhibit the strongest autopeak at 1 642 cm^-1, accompanied by the weakest negative cross peak at 1 006, 1 642 cm^-1, while Aksoo Apples display only one prominent autopeak at 1 010 cm^-1. Qin Guan Apples exhibit three autopeaks, whereas both 2001 Apples and New Century Apples have their strongest autopeak shifted to 1 020 cm^-1, differing by 10 wavenumbers from those of other varieties. This 2D-IR analysis enables differentiation between certain apple samples based on their unique spectral signatures. An optimized approach to analyzing the spectra of 216 apples from eight varieties utilizing Convolutional Neural Network (CNN) and Radial Basis Function (RBF) neural networks. By randomly selecting 152 sample spectra for model training and through iterative refinement and rigorous training, both models achieved an optimal state, attaining a classification accuracy of 100% on the training set. Following this, predictions were conducted on the spectra of an additional 64 samples, yielding remarkable results: an accuracy rate of 89.06% in CNN analysis and an even higher 90.6% in RBF neural network analysis. Both neural network models have demonstrated outstanding performance in terms of classification accuracy. Hence, the integration of FTIR, 2D-IR, CNN, and RBF neural network analysis methods forms a complementary approach in the study of apple analysis and identification, facilitating precise classification of Zhaotong apples. Additionally, this comprehensive methodology holds significant potential for application in the classification and identification of diverse substances beyond apples.

Key words：Zhaotong apples; Fourier transform infrared spectroscopy (FTIR); Two-dimensional infrared correlation spectroscopy (2D-IR); Neural network analysis model; Identification

Received: 2024-08-27 Accepted: 2024-12-17

ZTFLH:

O433.4

Cite this article:

MA Dian-xu,CAI Yan,LI Xiao-pan, et al. Comprehensive Identification and In-Depth Analysis of Zhaotong Apples Facilitated by Multidimensional Fourier Transform Infrared Spectroscopy Integrated With Two Neural Network Models[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(06): 1543-1550.

URL:

https://www.gpxygpfx.com/EN/10.3964/j.issn.1000-0593(2025)06-1543-08 OR https://www.gpxygpfx.com/EN/Y2025/V45/I06/1543

[1] WANG Chen-chen, ZHAI Ming-can, LI He, et al(王晨晨, 翟明灿, 李贺, 等). Food and Machinery(食品与机械), 2024, 40(7): 117.
[2] LIU Yan, FAN Yu, LIU Jiang, et al(刘炎, 樊玉, 刘江, 等). Environmental Science(环境科学), 2024，45(9): 5538.
[3] YI Li-mi-nu-er, HE Miao, LU Biao, et al(伊丽米努尔, 何苗, 陆彪, 等). Food Science(食品科学), 2020, 41(19): 62.
[4] PAN Xin-long, KONG Bao-hua, LI Yun-guo, et al(潘新龙, 孔宝华, 李云国, 等). Southern Fruit Trees of China(中国南方果树), 2021, 50(2): 151.
[5] Feng S, Yi J, Li X, et al. Journal of Agricultural and Food Chemistry, 2021, 69(1): 7.
[6] Srednicka-Tober D, Barański M, Kazimierczak R, et al. Applied Sciences, 2020, 10(9): 2997.
[7] Uselis N, Viškelis J, Lanauskas J, et al. Zemdirbyste-Agriculture，2020，107(4): 367.
[8] Zhang W, Pan Y, Jiang Y, et al. Critical Reviews in Food Science and Nutrition，2024, 64(24): 8689.
[9] Janne Lempe, Henryk Flachowsky, Andreas Peil. Physiologia Plantarum，2022; 174(5): e13782.
[10] Radu E Sestraş, Adriana F Sestraş. Plants，2023，12(4): 903.
[11] Wolkers W F, Oldenhof H. Methods in Molecular Biology, 2021, 2180: 3.
[12] Ribeiro D C S Z, Neto H A, Lima J S, et al. Heliyon, 2023, 9(1): e12898.
[13] Li Dandan, Zhang Dongyan, Jiang Dequan, et al. Ukrainian Journal of Physical Optics, 2022，23(4): 267.
[14] Christabel Y E Tachie, Daniel Obiri-Ananey, Marcela Alfaro-Córdoba, et al. Food Chemistry，2024，431: 137077.
[15] Shyam Narayan Jha, Pranita Jaiswal, Leena Kumari, et al. Agricultural Research，2021，10(2): 314.
[16] DAI Yuan-feng, DAI Zuo-xiao, GUO Guang-zhi, et al(戴元丰, 代作晓, 郭光智, 等). Journal of Chinese Institute of Food Science and Technology(中国食品学报), 2023, 23(9): 331.
[17] Liang Haibo, Chen Haifeng, Guo Jinhong, et al. IEEE Transactions on Instrumentation and Measurement，2022，71: 2504510.
[18] CHEN Feng-xia, YANG Tian-wei, LI Jie-qing, et al(陈凤霞，杨天伟，李杰庆, 等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2020, 40(12): 3839.