Rapid Detection of Tea Polyphenols in Fresh Tea Leaves Using Visible/Shortwave and Longwave Near-Infrared Spectroscopy

doi:10.12133/j.smartag.SA202505034

Abstract

Abstract:

[Objective] Tea polyphenols, as a key indicator for evaluating tea quality, possess significant health benefits. Traditional detection methods suffer from poor timeliness, high cost, and destructive sampling, making them difficult to meet the demands of tea cultivar breeding and real-time monitoring of tea quality. Meanwhile, rapid identification of tea cultivars and leaf positions is of great significance for guiding tea production. Therefore, the aim is to develop a non-destructive detection device for quality components of fresh tea leaves based on the combined technology of visible/short-wave near-infrared and long-wave near-infrared spectroscopy, to realize rapid non-destructive detection of tea polyphenol content and rapid identification of tea cultivars and leaf positions. [Methods] A rapid non-destructive detection device for quality components of fresh tea leaves was developed by combining visible/short-wave near-infrared spectroscopy (400-1 050 nm) and long-wave near-infrared spectroscopy (1 051-1 650 nm). The Savitzky-Golay (SG) convolution smoothing method was used for preprocessing the spectral data. The Folin-Ciocalteu method was employed to determine the tea polyphenol content, and abnormal samples were eliminated using the interquartile range (IQR) method. Data-level and feature-level fusion methods were adopted, with the competitive adaptive reweighted sampling (CARS) algorithm used to extract characteristic wavelengths. Prior to modeling, the Kennard-Stone algorithm was applied to partition the dataset into a training set and a prediction set at a ratio of 4:1. Models such as principal component analysis (PCA), partial least squares-discriminant analysis (PLS-DA), least squares support vector machine (LS-SVM), random forest (RF), extreme learning machine (ELM), and 1D convolutional neural network (1D-CNN) were constructed for the identification of 3 cultivars (Huangdan, Tieguanyin, and Benshan) and 4 leaf positions. Additionally, models including partial least squares regression (PLSR), least squares support vector regression (LS-SVR), RF, ELM, and 1D-CNN were established for predicting the tea polyphenol content in fresh tea leaves. [Results and Discussions] The results showed that there were significant differences in tea polyphenol contents among different cultivars and leaf positions (P<0.05). Specifically, the tea polyphenol content of Huangdan was (17.54±1.82%), which was 1.16 times and 1.04 times that of Tieguanyin (15.04±1.22%) and Benshan (16.81±1.24%), respectively. For each tea cultivar, the tea polyphenol content generally showed a decreasing trend from the 1st to the 4th leaf position, with the highest content in the 1st leaf position. Principal component analysis (PCA) revealed that for cultivar identification, the scatter distribution of the principal components of Huangdan, Tieguanyin, and Benshan, as well as their projections in the directions of PC1 and PC2, showed a clear trend of clustering into three groups, indicating a good classification effect, although there was still some overlap among individual samples. For leaf position identification, the scatter distributions of the principal components of the 1st, 2nd, 3rd, and 4th leaf positions overlapped with each other, with no obvious clustering among leaf positions. Compared with single-source data, models based on data fusion effectively improved prediction performance. Among them, the PLS-DA model established by combining SG preprocessing with feature-level fusion achieved prediction accuracies of 100% and 87.93% for the identification of 3 tea cultivars and 4 leaf positions, respectively. Furthermore, the 1D-CNN model based on data-level fusion exhibited superior performance in predicting tea polyphenol content, with a coefficient of determination (R ² _P), root mean square error of prediction (RMSEP), and residual predictive deviation (RPD) of 0.802 0, 0.636 8%, and 2.268 4, respectively, which outperformed models using only visible/short-wave near-infrared spectroscopy or long-wave near-infrared spectroscopy. [Conclusions] The developed detection device combining visible/short-wave near-infrared and long-wave near-infrared spectroscopy, mainly composed of spectrometers, Y-type optical fibers, plant probes, polymer lithium batteries, DC uninterruptible power supplies, voltage conversion modules, and aluminum alloy casings, could synchronously collect multi-source spectral data of visible/short-wave near-infrared and long-wave near-infrared from fresh tea leaves. Combined with data fusion methods and machine learning algorithms, it enabled rapid detection of tea polyphenol content and efficient identification of cultivars and leaf positions in fresh tea leaves, providing new insights for the application of multi-source data fusion technology in elite tea cultivar breeding and non-destructive detection of fresh tea leaf quality.

Key words: Fresh tea leaves, tea polyphenols, non-destructive detection, data fusion, one-dimensional convolutional neural network (1D-CNN)

CLC Number:

XU Jinchai, LI Xiaoli, WENG Haiyong, HE Yong, ZHU Xuesong, LIU Hongfei, HUANG Zhenxiong, YE Dapeng. Rapid Detection of Tea Polyphenols in Fresh Tea Leaves Using Visible/Shortwave and Longwave Near-Infrared Spectroscopy[J]. Smart Agriculture, doi: 10.12133/j.smartag.SA202505034.

Figures/Tables 13

Fig. 1

Fig.2

Fig. 3

Fig.4

Fig.5

Fig.6

Fig.7

Fig.8

Table 1

Comparison of performance between single-sensor and multi-source data fusion models for tea variety and leaf position identification

类别	识别模型	输入变量	变量数	训练集				预测集
类别	识别模型	输入变量	变量数	准确率/%	精确率%	召回率/%	F ₁分数	准确率/%	精确率%	召回率/%	F ₁分数
3个品种	PLS-DA	可见/短波近红外	651	100.00	100.00	100.00	1.000 0	100.00	100.00	100.00	1.000 0
		长波近红外	600	87.83	87.94	88.17	0.879 9	94.83	94.11	93.70	0.938 5
		数据级数据融合	1 251	100.00	100.00	100.00	1.000 0	100.00	100.00	100.00	1.000 0
		特征级数据融合	171	99.13	99.12	99.10	0.991 1	100.00	100.00	100.00	1.000 0
	LS-SVM	可见/短波近红外	651	99.57	99.55	99.55	0.995 5	100.00	100.00	100.00	1.000 0
		长波近红外	600	96.52	96.66	96.60	0.966 1	100.00	100.00	100.00	1.000 0
		数据级数据融合	1 251	99.57	99.58	99.54	0.995 6	100.00	100.00	100.00	1.000 0
		特征级数据融合	171	98.26	98.29	98.22	0.982 1	100.00	100.00	100.00	1.000 0
	RF	可见/短波近红外	651	100.00	100.00	100.00	1.000 0	100.00	100.00	100.00	1.000 0
		长波近红外	600	100.00	100.00	100.00	1.000 0	98.28	98.15	97.92	0.980 3
		数据级数据融合	1 251	100.00	100.00	100.00	1.000 0	100.00	100.00	100.00	1.000 0
		特征级数据融合	171	100.00	100.00	100.00	1.000 0	100.00	100.00	100.00	1.000 0
	ELM	可见/短波近红外	651	100.00	100.00	100.00	1.000 0	100.00	100.00	100.00	1.000 0
		长波近红外	600	100.00	100.00	100.00	1.000 0	100.00	100.00	100.00	1.000 0
		数据级数据融合	1 251	100.00	100.00	100.00	1.000 0	100.00	100.00	100.00	1.000 0
		特征级数据融合	171	98.70	98.69	98.69	0.986 9	100.00	100.00	100.00	1.000 0
	1D-CNN	可见/短波近红外	651	100.00	100.00	100.00	1.000 0	100.00	100.00	100.00	1.000 0
		长波近红外	600	93.91	93.99	94.02	0.939 8	98.28	98.15	97.92	0.979 7
		数据级数据融合	1 251	100.00	100.00	100.00	1.000 0	100.00	100.00	100.00	1.000 0
		特征级数据融合	171	100.00	100.00	100.00	1.000 0	100.00	100.00	100.00	1.000 0
4个叶位	PLS-DA	可见/短波近红外	651	90.87	91.24	90.93	0.909 8	79.31	82.39	79.76	0.804 2
		长波近红外	600	64.35	64.92	64.39	0.644 8	68.97	69.02	69.74	0.689 7
		数据级数据融合	1 251	94.35	94.45	94.40	0.943 9	84.48	86.06	86.87	0.858 2
		特征级数据融合	90	85.65	85.45	85.23	0.853 2	87.93	91.18	88.51	0.889 3
	LS-SVM	可见/短波近红外	651	73.91	74.08	74.31	0.734 1	58.62	58.41	61.62	0.577 0
		长波近红外	600	56.09	56.00	56.57	0.560 9	58.62	58.86	59.93	0.592 1
		数据级数据融合	1 251	80.87	81.00	81.33	0.805 2	74.14	73.42	72.00	0.722 7
		特征级数据融合	90	96.52	96.38	96.46	0.964 1	79.31	81.55	78.57	0.789 3
	RF	可见/短波近红外	651	96.52	96.52	96.56	0.965 4	74.14	72.92	72.79	0.728 5
		长波近红外	600	97.83	97.84	97.82	0.978 3	75.86	75.86	76.94	0.764 0
		数据级数据融合	1 251	100.00	100.00	100.00	1.000 0	79.31	79.49	78.73	0.791 1
		特征级数据融合	90	79.57	79.27	79.69	0.794 8	77.59	80.71	78.78	0.797 4
	ELM	可见/短波近红外	651	95.22	95.24	95.48	0.953 6	75.86	75.45	76.43	0.759 3
		长波近红外	600	65.65	65.57	65.07	0.653 2	67.24	68.20	71.36	0.697 5
		数据级数据融合	1 251	79.57	79.49	79.43	0.794 6	79.31	80.33	78.99	0.796 5
		特征级数据融合	90	86.09	85.64	85.97	0.858 1	74.14	79.29	76.95	0.781 0
	1D-CNN	可见/短波近红外	651	79.57	79.64	79.47	0.794 5	68.97	68.23	69.10	0.684 6
		长波近红外	600	62.17	62.29	63.57	0.630 0	63.79	63.29	64.64	0.638 2
		数据级数据融合	1 251	87.83	87.96	88.28	0.876 2	74.14	75.09	73.62	0.735 8
		特征级数据融合	90	97.39	97.37	97.34	0.973 3	82.76	84.32	86.06	0.842 1

Table 1

Fig.9

Table 2

Table 3

Fig. 10

References 30

[1]	SUN M F , JIANG C L , KONG Y S , et al . Recent advances in analytical methods for determination of polyphenols in tea: A comprehensive review[J]. Foods, 2022, 11(10): ID 1425.
[2]	CHEN G C , YI Z , CHEN X Y , et al . Polyphenol nanoparticles from commonly consumed tea for scavenging free radicals, stabilizing pickering emulsions, and inhibiting cancer cells[J]. ACS applied nano materials, 2020, 4(1), 652-665.
[3]	JANG M , PARK R , PARK Y I , et al . EGCG, a green tea polyphenol, inhibits human coronavirus replication in vitro [J]. Biochemical and biophysical research communications, 2021, 547: 23-28.
[4]	XING L J , ZHANG H , QI R L , et al . Recent advances in the understanding of the health benefits and molecular mechanisms associated with green tea polyphenols[J]. Journal of agricultural and food chemistry, 2019, 67(4): 1029-1043.
[5]	MATSUSHITA K , HONDA C , NAKAMURA Y , et al . Comparison of colorimetric methods for the analysis of total polyphenols in green tea extracts[J]. Bioscience, biotechnology, and biochemistry, 2024, 88(7): 798-803.
[6]	GHARIBZAHEDI S M T , BARBA F J , ZHOU J J , et al . Electronic sensor technologies in monitoring quality of tea: A review[J]. Biosensors, 2022, 12(5): ID 356.
[7]	HUANG Y F , DONG W T , SANAEIFAR A , et al . Development of simple identification models for four main catechins and caffeine in fresh green tea leaf based on visible and near-infrared spectroscopy[J]. Computers and electronics in agriculture, 2020, 173: ID 105388.
[8]	REN G X , LIU Y , NING J M , et al . Assessing black tea quality based on visible–near infrared spectra and kernel-based methods[J]. Journal of food composition and analysis, 2021, 98: ID 103810.
[9]	BIANCOLILLO A , MARINI F . Special issue "application of spectroscopy in food analysis"[J]. Applied sciences, 2021, 11(9): ID 3860.
[10]	CHEN S M , WANG C Y , TSAI C Y , et al . Fermentation quality evaluation of tea by estimating total catechins and theanine using near-infrared spectroscopy[J]. Vibrational spectroscopy, 2021, 115: ID 103278.
[11]	SANAEIFAR A , HUANG X Y , CHEN M Y , et al . Nondestructive monitoring of polyphenols and caffeine during green tea processing using Vis-NIR spectroscopy[J]. Food science & nutrition, 2020, 8(11): 5860-5874.
[12]	李晓丽, 张东毅, 董雨伦, 等 . 基于卷积神经网络的茶鲜叶主要内含物的光谱快速检测方法[J]. 中国农业大学学报, 2021, 26(11): 113-122.
	LI X L , ZHANG D Y , DONG Y L , et al . Spectral rapid detection of phytochemicals in tea(Camellia sinensis)based on convolutional neural network[J]. Journal of China agricultural university, 2021, 26(11): 113-122.
[13]	WANG X Y , CHEN H C , JI R D , et al . Detection of carmine in black tea based on UV-vis absorption spectroscopy and machine learning[J]. Food analytical methods, 2025, 18(2): 149-160.
[14]	HERSHBERGER J , MBANJO E G N , PETETI P , et al . Low-cost, handheld near-infrared spectroscopy for root dry matter content prediction in cassava[J]. The plant phenome journal, 2022, 5(1): ID e20040.
[15]	王凡, 赵春江, 徐波, 等 . 便携式茶鲜叶品质光谱检测装置研制[J]. 农业工程学报, 2020, 36(24): 273-280.
	WANG F , ZHAO C J , XU B , et al . Development of a portable detection device for the quality of fresh tea leaves using spectral technology[J]. Transactions of the Chinese society of agricultural engineering, 2020, 36(24): 273-280.
[16]	李文萃, 周新奇, 范起业, 等 . 便携式近红外茶叶品质快速检测仪设计与试验[J]. 现代食品科技, 2021, 37(5): 303-309.
	LI W C , ZHOU X Q , FAN Q Y , et al . Design and experiment of portable near-infrared tea quality detector[J]. Modern food science and technology, 2021, 37(5): 303-309.
[17]	WANG W X , PENG Y K , SUN H W , et al . Real-time inspection of pork quality attributes using dual-band spectroscopy[J]. Journal of food engineering, 2018, 237: 103-109.
[18]	RYCKEWAERT M , CHAIX G , HÉRAN D , et al . Evaluation of a combination of NIR micro-spectrometers to predict chemical properties of sugarcane forage using a multi-block approach[J]. Biosystems engineering, 2022, 217: 18-25.
[19]	董春旺, 刘中原, 杨明, 等 . 基于多源信息融合的绿茶杀青叶水分含量智能感知方法[J]. 食品科学, 2022, 43(20): 242-251.
	DONG C W , LIU Z Y , YANG M , et al . Intelligent sensing method for detecting moisture content in fixed tea leaves for green tea based on multi-source information fusion[J]. Food science, 2022, 43(20): 242-251.
[20]	XU J C , QU F F , SHEN B H , et al . Rapid detection of tea polyphenols in fresh tea leaves based on fusion of visible/short-wave and long-wave near infrared spectroscopy and its device development[J]. Applied sciences, 2023, 13(3): ID 1739.
[21]	BANAS K , BANAS A M , PASTORIN G , et al . Sensing the changes in stratum corneum using Fourier transform infrared microspectroscopy and hyperspectral data processing[J]. Sensors, 2024, 24(21): ID 7054.
[22]	ZHANG G S , HAO H , WANG Y C , et al . Optimized adaptive Savitzky-Golay filtering algorithm based on deep learning network for absorption spectroscopy[J]. Spectrochimica acta part A: Molecular and biomolecular spectroscopy, 2021, 263: ID 120187.
[23]	WANG S H , ZHANG M Y , ZHAO Z X , et al . Optical properties and Monte Carlo simulation of cotton fibers with different micronaire values[J]. Industrial crops and products, 2025, 223: ID 120254.
[24]	饶利波, 陈晓燕, 庞涛 . 基于光谱技术的Bipls算法结合CARS算法的苹果可溶性固形物含量检测[J]. 发光学报, 2019, 40(3): ID 389.
	RAO L B , CHEN X Y , PANG T . Determination of apple soluble solids content using bipls coupled with CARS algorithm based on spectral technology[J]. Chinese journal of luminescence, 2019, 40(3): ID 389.
[25]	翁海勇, 许金钗, 陶铸, 等 . 高EGCG含量茶树品种光谱识别模型构建[J]. 中国农机化学报, 2021, 42(6): 111-117.
	WENG H Y , XU J C , TAO Z , et al . Construction of spectral screening model for tea cultivars with high EGCG content[J]. Journal of Chinese agricultural mechanization, 2021, 42(6): 111-117.
[26]	安琪, 王占彬, 安国庆, 等 . 基于随机森林-遗传算法-极限学习机的非侵入式负荷识别方法[J]. 科学技术与工程, 2022, 22(5): 1929-1935.
	AN Q , WANG Z B , AN G Q , et al . Non-intrusive load identification method based on RF-GA-ELM[J]. Science technology and engineering, 2022, 22(5): 1929-1935.
[27]	YANG B Y , YANG Z L , XU Y , et al . A 1D-CNN model for the early detection of Citrus Huanglongbing disease in the sieve plate of phloem tissue using micro-FTIR[J]. Chemometrics and intelligent laboratory systems, 2024, 252: ID 105202.
[28]	ZHOU R Q , LI X L , HE Y , et al . Determination of catechins and caffeine content in tea (Camellia sinensis L.) leaves at different positions by Fourier-transform infrared spectroscopy[J]. Transactions of the asabe, 2018, 61(4): 1221-1230.
[29]	李晓丽, 魏玉震, 徐劼, 等 . 基于高光谱成像的茶叶中EGCG分布可视化[J]. 农业工程学报, 2018, 34(7): 180-186.
	LI X L , WEI Y Z , XU J , et al . EGCG distribution visualization in tea leaves based on hyperspectral imaging technology[J]. Transactions of the Chinese society of agricultural engineering, 2018, 34(7): 180-186.
[30]	姚奉奇, 陶骏骏, 王海晖, 等 . 茶多酚热解特性及其反应机理研究[J]. 林产化学与工业, 2017, 37(5): 19-27.
	YAO F Q , TAO J J , WANG H H , et al . Study of pyrolysis behavior and reaction mechanism of tea polyphenols[J]. Chemistry and industry of forest products, 2017, 37(5): 19-27.

输入变量	数据集	样本数/个	最小值/%	最大值/%	平均值/%	标准差/%	变异系数/%	偏度	峰度
可见/短波近红外（400~1 050 nm）	全集	276	12.71	20.60	16.41	1.60	9.80	0.019	-0.450
	训练集	221	12.71	20.60	16.44	1.62	9.86	0.036	-0.472
	预测集	55	13.13	19.58	16.25	1.55	9.59	-0.089	-0.352
长波近红外（1 051~1 650 nm）	全集	276	12.71	20.60	16.41	1.60	9.80	0.019	-0.450
	训练集	221	12.71	20.60	16.47	1.67	10.17	0.020	-0.571
	预测集	55	13.18	18.36	16.16	1.28	7.97	-0.402	-0.273
数据级数据融合	全集	276	12.71	20.60	16.41	1.60	9.80	0.019	-0.450
	训练集	221	12.71	20.60	16.48	1.64	9.95	0.037	-0.472
	预测集	55	13.18	18.72	16.09	1.44	8.97	-0.288	-0.748
特征级数据融合	全集	276	12.71	20.60	16.41	1.60	9.80	0.019	-0.450
	训练集	221	12.71	20.60	16.51	1.62	9.82	0.064	-0.484
	预测集	55	13.13	18.29	15.98	1.49	9.34	-0.372	-0.855

定量模型	输入变量	变量数/个	R ² _C	RMSEC/%	R ² _P	RMSEP/%	RPD
PLSR	可见/短波近红外	651	0.729 5	0.841 5	0.787 2	0.712 3	2.167 8
	长波近红外	600	0.461 4	1.227 0	0.465 6	0.933 1	1.368 0
	数据级数据融合	1 251	0.743 7	0.828 5	0.789 2	0.657 0	2.178 5
	特征级数据融合	106	0.700 1	0.886 2	0.770 2	0.708 9	2.086 3
LS-SVR	可见/短波近红外	651	0.650 4	0.956 7	0.738 3	0.789 8	1.955 0
	长波近红外	600	0.550 4	1.121 0	0.587 0	0.820 3	1.556 1
	数据级数据融合	1 251	0.741 2	0.832 5	0.751 3	0.713 7	2.005 5
	特征级数据融合	106	0.755 7	0.799 9	0.743 6	0.748 7	1.975 2
RF	可见/短波近红外	651	0.928 4	0.432 7	0.751 7	0.769 4	2.007 0
	长波近红外	600	0.910 6	0.499 7	0.591 2	0.816 2	1.564 1
	数据级数据融合	1 251	0.929 4	0.434 7	0.768 5	0.688 5	2.078 8
	特征级数据融合	106	0.886 2	0.545 8	0.740 7	0.753 1	1.963 9
ELM	可见/短波近红外	651	0.637 7	0.973 9	0.745 6	0.778 7	1.982 8
	长波近红外	600	0.633 4	1.012 2	0.664 5	0.739 4	1.726 6
	数据级数据融合	1 251	0.647 3	0.971 9	0.753 2	0.710 9	2.013 3
	特征级数据融合	106	0.768 3	0.778 9	0.787 0	0.682 5	2.166 8
1D-CNN	可见/短波近红外	651	0.836 3	0.654 5	0.724 2	0.810 8	1.921 9
	长波近红外	600	0.534 0	1.141 2	0.622 0	0.784 4	1.642 4
	数据级数据融合	1 251	0.842 5	0.649 4	0.802 0	0.636 8	2.268 4
	特征级数据融合	106	0.819 6	0.687 3	0.733 4	0.763 5	1.954 8