WANG Gui-yao,LIN Meng-han²,LI Shao-peng³,ZHAN Ying⁴,ZHANG Jun⁴,PENG Yun-fa⁴,TIAN Zhen²,ZHOU Han-ping,GUO Jian-hua,SONG Ji-zhen.Determination of Nicotine in Tobacco by NIR Spectroscopy Enhanced with Deep Learning and Kepler Optimization Algorithm[J].Journal of Instrumental Analysis,2025,44(10):2071-2078.
WANG Gui-yao,LIN Meng-han²,LI Shao-peng³,ZHAN Ying⁴,ZHANG Jun⁴,PENG Yun-fa⁴,TIAN Zhen²,ZHOU Han-ping,GUO Jian-hua,SONG Ji-zhen.Determination of Nicotine in Tobacco by NIR Spectroscopy Enhanced with Deep Learning and Kepler Optimization Algorithm[J].Journal of Instrumental Analysis,2025,44(10):2071-2078. DOI: 10.12452/j.fxcsxb.250425320.
Determination of Nicotine in Tobacco by NIR Spectroscopy Enhanced with Deep Learning and Kepler Optimization Algorithm
This study aims to improve the quantitative analysis accuracy of nicotine content in tobacco leav
es using near-infrared(NIR) spectroscopy by proposing a deep learning model integrating the Kepler optimization algorithm(KOA),convolutional neural network(CNN),gated recurrent unit(GRU),and multi-head self-attention mechanism(MultiAttention). First,the NIR spectra of 1 790 tobacco leaf samples were preprocessed using the Savitzky-Golay first derivative method. The CNN was employed to extract multi-scale spectral features,the GRU was used to capture the sequential dependencies between wavelength points,and the MultiAttention mechanism was introduced to dynamically weight features. Additionally,the KOA algorithm was applied to optimize the model's hyperparameters (learning rate,number of convolutional kernels,and hidden layer nodes). By adjusting the orbital period factor(
T
C
),initial gravitational strength(
M
₀),and decay coefficient(
λ
),the model addressed the issues of traditional models being prone to local optima and slow convergence. Experimental results showed that when
T
C
=1,
M
₀=0.05,and
λ
=8,the model achieved a goodness-of-fit(
R
²) of 0.980,root mean square error(RMSE) of 0.069,and mean absolute error(MAE) of 0.049 for nicotine content prediction,significantly outperforming comparative models such as partial least squares(PLS) and convolutional neural network regression(CNNR). The study demonstrates that this model,through an integrated framework of feature extraction-sequential modeling-global optimization,effectively enhances the robustness and generalization ability of NIR spectral quantitative analysis,providing a new method for rapid and accurate detection of chemical components in tobacco leaves.
关键词
Keywords
references
Zhang X B , Xu Z Q , Zhong Y J , Zhu H F , Li Z , Zhang J , Zhan Y , Peng Y F , Liu J G . J. Instrum. Anal. (张晓兵,徐志强,钟永健,朱宏福,李峥,张军,詹映,彭云发,刘建国. 分析测试学报), 2024 , 43 ( 5 ): 792 - 797 .
Wu J Z , Shi Y D , Huang H , Li X R . J. Instrum. Anal. (吴继忠,时艺丹,黄慧,厉小润. 分析测试学报), 2023 , 42 ( 9 ): 1112 - 1118 .
Chu X L , Chen P , Li J Y , Liu D , Xu Y P . J. Instrum. Anal. (褚小立,陈瀑,李敬岩,刘丹,许育鹏. 分析测试学报), 2020 , 39 ( 10 ): 1181 - 1188 .
Zong Q Q , Ding X Q , Han F , Gong H L , Zhang L . Comput. Digital Eng. (宗倩倩,丁香乾,韩凤,宫会丽,张磊. 计算机与数字工程), 2019 , 47 ( 2 ): 275 - 280 .
Wang C , Wu X H , Li L Q , Wang Y S , Li Z W . Spectrosc. Spectral Anal. (王璨,武新慧,李恋卿,王玉顺,李志伟. 光谱学与光谱分析), 2018 , 38 ( 1 ): 36 - 41 .
Liu Z Y , Zhang C H , Jiang J K , Shen B G , Ding Y F , Zhang L L , Zhu C . Spectrosc. Spectral Anal. (刘宗溢,张彩虹,蒋健康,沈斌国,丁艳菲,张雷蕾,朱诚. 光谱学与光谱分析), 2024 , 44 ( 4 ): 1018 - 1024 .
Xu S Y , Liu Z Y , Huang Y , Zeng Y , Bie Z L , Dong W J . Trans. Chin. Soc. Agric. Mach .(徐 胜勇,刘政义,黄远,曾雨,别之龙,董万静. 农业机械学报), 2024 , 55 ( 8 ): 243 - 252 .
Jiang H , Wang Y G , Chen J , Huang X Y . Acta Opt . Sin. (江灏,王尤刚,陈静,黄新宇. 光学学报), 2020 , 40 ( 7 ): 12 - 19 .
Li H , Zhao Q , Cui C Z , Fan D W , Zhang C K , Shi Y C , Wang Y . Spectrosc. Spectral Anal. (李浩,赵青,崔辰州,樊东卫,张成奎,史艳翠,王嫄. 光谱学与光谱分析), 2024 , 44 ( 6 ): 1668 - 1675 .
Miao J F , Tang B , Chen Q , Long Z R , Ye B Q , Zhou Y , Zhang J F , Zhao M F , Zhou M . J. Atmos. Environ. Opt . (缪俊锋,汤斌,陈庆,龙邹荣,叶彬强,周彦,张金富,赵明富,周密. 大气与环境光学学报), 2024 , 19 ( 1 ): 73 - 84 .
Wu J G , Wang H , Zhang L , Yang L S , Peng J J , Gong M C . Environ. Sci. (吴建高,汪泓,张磊,杨隆珊,彭俊杰,龚明冲 . 环境科学) , 2025 , 46 ( 4 ): 2313 - 2324 .
Guan X M , Cui H B . Sci. Technol. Eng. (管雪梅,崔宏博. 科学技术与工程), 2024 , 24 ( 28 ): 12268 - 12276 .
He S D , Li J , Hu C J , Wang J Q , Zhang Y Y , Gao Z Y , Du Z H , Cao Y . J. Gun Launch & Control(何淑典 , 李杰 , 胡陈君 , 王镜淇 , 张元园 , 高正阳 , 杜增辉 , 曹宇 . 火炮发射与控制学报) , 2025 , 46 ( 3 ): 18 - 24 .
Wang M Y , Ma X J , Ge L J , Zhou S H , Xu Z W , Wu H , Ren X D . J . Inner Mongolia Agric. Univ. : Nat. Sci. Ed .
YC/T 468 - 2021 . Determination of Total Alkaloids in Tobacco and Tobacco Products-Continuous Flow (Potassium Thiocyanate) Method. Tobacco Industry Standard of the People's Republic of China(烟草及烟草制品 总植物碱的测定 连续流动(硫氰酸钾)法. 中华人民共和国烟草行业标准) .