Constructing Rapid and Undamaged Detection Models for Main Physicochemical Indexes of Brewing Sorghum Based on Near Infrared Spectrum
-
摘要: 为满足对于酒用高粱直链淀粉、支链淀粉、蛋白质、脂肪、单宁含量快速检测的需求,本文采用17种光谱数据预处理方法和4种波段挑选算法建立了这些指标的近红外光谱分析模型。结果表明,各指标最佳光谱预处理方法分别为一阶导数+多元散射校正+Z-score标准化、矢量归一化+均指中心化、标准正态变量变换+Z-score标准化、多元散射校正、标准正态变量变换+Z-score标准化,预测直链淀粉、支链淀粉、蛋白质、单宁含量最佳的波段挑选方法为蒙特卡洛-无信息变量消除,脂肪为竞争自适应重加权采样法。整粒高粱这5项指标最优模型的决定系数(R2)分别为0.9560、0.8765、0.9069、0.8658、0.8841,交叉验证均方根误差(RMSECV)值分别为1.3222、2.3477、0.3549、0.2164、0.1077,外部独立样品验证结果显示模型预测准确率高。本文所建立的近红外分析模型可为酿酒行业实现对高粱的快检提供技术参考。Abstract: To satisfy the demands of rapid determination of amylose, amylopectin, protein, fat, and tannin contents in brewing sorghums, in this paper, 17 spectral data preprocessing methods and 4 wavelength band selection algorithms were used to establish the near infrared spectral analysis models for these indexes. The results showed that the best spectral preprocessing methods for each index were 1st der (1st)+multiplicative scatter correction (MSC)+Z-score standardization (ZS), vector normalization (VN)+mean centering (MC), standard normal variate transformation (SNV)+ZS, MSC, SNV+ZS, respectively. The best wavelength band selection algorithm for predicting amylose, amylopectin, protein, and tannin contents was monte-carlo uninformative variable elimination, and that of fat was competitive adaptive reweighted sampling. The R2 in the optimal models for these 5 indexes of whole grain sorghums were 0.9560, 0.8765, 0.9069, 0.8658, 0.8841, and the RMSECV values were 1.3222, 2.3477, 0.3549, 0.2164, 0.1077, respectively. The validation results of external independent samples showed that the models had a high prediction accuracy. The NIR analysis model established in this study could provide a technical reference for the rapid detection of sorghums in the brewing industry.
-
表 1 不同光谱预处理方法高粱理化指标的主要评价参数
Table 1. Main evaluation parameters of physical and chemical indexes of sorghum with different spectrum pretreatments
建模指标 序号 预处理方法 附加预处理方法 波长变量 推荐因子数 RMSECV 直链淀粉 1 无 无 780 19 2.2603 2 无 均值中心化 780 19 2.1086 3 无 Z-score标准化 780 19 2.0766 4 标准正态变量变换+去趋势 Z-score标准化 780 17 2.0760 5 去趋势 Z-score标准化 780 18 2.0918 6 减去一条直线 Z-score标准化 780 17 2.1389 7 多元散射校正 Z-score标准化 780 16 2.1504 8 一阶导+多元散射校正 Z-score标准化 780 15 2.2052 9 标准正态变量变换 Z-score标准化 780 17 2.2627 10 矢量归一化 Z-score标准化 780 18 2.3058 支链淀粉 1 无 无 780 14 3.3035 2 无 均值中心化 780 13 3.1574 3 无 Z-score标准化 780 13 3.1807 4 矢量归一化 均值中心化 780 16 2.7791 5 一阶导+多元散射校正 均值中心化 780 10 2.8253 6 一阶导+标准正态变量变换 均值中心化 780 10 2.8503 7 一阶导+矢量归一化 均值中心化 780 10 2.8767 8 多元散射校正 均值中心化 780 12 2.9310 9 减去一条直线 均值中心化 780 12 2.9604 10 一阶导+减去一条直线 均值中心化 780 10 2.9787 蛋白质 1 无 无 780 10 0.5933 2 无 均值中心化 780 13 0.5803 3 无 Z-score标准化 780 13 0.5732 4 标准正态变量变换+去趋势 Z-score标准化 780 18 0.4620 5 标准正态变量变换 Z-score标准化 780 19 0.4709 6 多元散射校正 Z-score标准化 780 18 0.4903 7 最大最小归一化 Z-score标准化 780 13 0.4933 8 一阶导+标准正态变量变换 Z-score标准化 780 6 0.4947 9 一阶导+减去一条直线 Z-score标准化 780 8 0.4957 10 一阶导+多元散射校正 Z-score标准化 780 6 0.5115 脂肪 1 无 无 780 7 0.3377 2 无 均值中心化 780 7 0.3186 3 无 Z-score标准化 780 7 0.3313 4 标准正态变量变换 均值中心化 780 6 0.3148 5 矢量归一化 均值中心化 780 6 0.3148 6 多元散射校正 均值中心化 780 6 0.3175 7 减去一条直线 均值中心化 780 7 0.3186 8 导数 均值中心化 780 6 0.3214 9 最大最小归一化 均值中心化 780 6 0.3239 10 一阶导+多元散射校正 均值中心化 780 6 0.3240 单宁 1 无 无 780 9 0.2400 2 无 均值中心化 780 9 0.2384 3 无 Z-score标准化 780 9 0.2368 4 减去一条直线 Z-score标准化 780 10 0.2107 5 一阶导+减去一条直线 Z-score标准化 780 9 0.2196 6 标准正态变量变换+去趋势 Z-score标准化 780 9 0.2212 7 标准正态变量变换 Z-score标准化 780 9 0.2235 8 去趋势 Z-score标准化 780 9 0.2245 9 多元散射校正 Z-score标准化 780 9 0.2262 10 矢量归一化 Z-score标准化 780 7 0.2292 表 2 高粱理化指标近红外预测模型的主要参数
Table 2. Main parameters of the NIR prediction model for physicochemical indexes of sorghums
指标 序号 波段挑选方法 挑选波段(cm−1) 光谱处理方法 波长变量 因子数 RMSECV R2 直链淀粉 1 反向区间PLS-区间分段数10 4598、4606、… 去趋势+Z-score标准化 546 17 1.4690 0.9457 2 无信息变量消除 4058、4066、… 减去一条直线+Z-score标准化 156 13 1.6136 0.9345 3 蒙特卡罗-无信息变量消除 4027、4035、… 一阶导+多元散射校正+Z-score标准化 220 14 1.3222 0.9560 4 竞争自适应重加权采样 4042、4050、… 一阶导+多元散射校正+Z-score标准化 43 13 0.9561 0.9770 支链淀粉 1 反向区间PLS-区间分段数10 4004、4011、… 矢量归一化+均值中心化 390 14 2.5854 0.8502 2 无信息变量消除 4035、4042、… 一阶导+标准正态变量变换+均值中心化 164 8 2.5727 0.8516 3 蒙特卡罗-无信息变量消除 4197、4413、… 矢量归一化+均值中心化 90 13 2.3477 0.8765 4 竞争自适应重加权采样 4081、4197、… 矢量归一化+均值中心化 24 12 2.0409 0.9067 蛋白质 1 反向区间PLS-区间分段数10 4004、4011、… 标准正态变量变换+Z-score标准化 546 20 0.3631 0.9025 2 无信息变量消除 4104、4112、… 标准正态变量变换+去趋势+Z-score标准化 118 14 0.3894 0.8879 3 蒙特卡罗-无信息变量消除 4135、4359、… 标准正态变量变换+Z-score标准化 230 15 0.3549 0.9069 4 竞争自适应重加权采样 4652、4937、… 标准正态变量变换+去趋势+Z-score标准化 30 14 0.2668 0.9474 脂肪 1 反向区间PLS-区间分段数10 4004、4011、… 矢量归一化+均值中心化 624 7 0.2834 0.6963 2 反向区间PLS-区间分段数20 5200、5207、… 一阶导+矢量归一化+Z-score标准化 234 11 0.2340 0.7235 3 反向区间PLS-区间分段数30 5254、5261、… 一阶导+矢量归一化+Z-score标准化 234 11 0.2337 0.7269 4 无信息变量消除 4011、4019、… 导数+均值中心化 197 4 0.3069 0.6222 5 蒙特卡罗-无信息变量消除 4343、4613、… 多元散射校正+均值中心化 60 5 0.2936 0.7036 6 蒙特卡罗-无信息变量消除 4027、4050、… 标准正态变量变换+Z-score标准化 210 6 0.2567 0.7704 7 竞争自适应重加权采样 4760、4775、… 一阶导+矢量归一化+均值中心化 43 4 0.2606 0.8405 8 竞争自适应重加权采样 4613、4629、… 标准正态变量变换+Z-score标准化 36 7 0.2164 0.8658 单宁 1 反向区间PLS-区间分段数10 4598、4606、… 去趋势+Z-score标准化 468 11 0.1546 0.8136 2 反向区间PLS-区间分段数20 3996-5192、… 矢量归一化+Z-score标准化 546 18 0.1288 0.8569 3 反向区间PLS-区间分段数30 3996-5007、… 消除常数偏移量+Z-score标准化 546 13 0.1256 0.8608 4 无信息变量消除 4096、4212、… Z-score标准化 239 11 0.1449 0.8188 5 蒙特卡罗-无信息变量消除 4212、4220、… 标准正态变量变换+去趋势+Z-score标准化 150 11 0.1438 0.8103 6 竞争自适应重加权采样 4189、4235、… 消除常数偏移量+Z-score标准化 40 7 0.1078 0.8975 7 无 6061~10000 消除常数偏移量+Z-score标准化 492 13 0.1468 0.8212 8 蒙特卡罗-无信息变量消除 4259、4266、… 多元散射校正 140 12 0.1077 0.8841 注:模型验证方式为内部交叉验证,验证抽样方法为K折系统抽样。 表 3 高粱籽粒整粒和粉碎样品的近红外模型影响
Table 3. Comparison of NIR models of intact and crushed sorghum samples
测定指标 高粱形态 波段挑选方法 光谱处理方法 波长变量 因子数 RMSECV R2 直链淀粉 整粒 蒙特卡罗-无信息变量消除 一阶导+多元散射校正+Z-score标准化 220 14 1.3222 0.9560 粉碎 蒙特卡罗-无信息变量消除 一阶导+矢量归一化+Z-score标准化 176 13 0.8739 0.9742 支链淀粉 整粒 蒙特卡罗-无信息变量消除 矢量归一化+均值中心化 90 13 2.3477 0.8765 粉碎 蒙特卡罗-无信息变量消除 标准正态变量变换+去趋势+均值中心化 125 15 2.2879 0.8780 蛋白质 整粒 蒙特卡罗-无信息变量消除 标准正态变量变换+Z-score标准化 230 15 0.3549 0.9069 粉碎 蒙特卡罗-无信息变量消除 一阶导+标准正态变量变换+均值中心化 230 13 0.1171 0.9889 脂肪 整粒 竞争自适应重加权采样 标准正态变量变换+Z-score标准化 36 7 0.2164 0.8658 粉碎 竞争自适应重加权采样 Z-score标准化 85 13 0.0932 0.8692 单宁 整粒 蒙特卡罗-无信息变量消除 多元散射校正 140 12 0.1077 0.8841 粉碎 反向区间PLS-区间分段数30 多元散射校正+Z-score标准化 312 12 0.0776 0.9414 -
[1] 钟敏, 张健. 原料糯高粱对酱香型白酒品质影响的研究现状[J]. 中国酿造,2022,41(1):32−36. [ZHONG M, ZHANG J. Research status of effect of raw glutinous sorghum on quality of sauce-flavor Baijiu[J]. Chinese Brewing,2022,41(1):32−36. doi: 10.11882/j.issn.0254-5071.2022.01.006 [2] 孙细珍, 熊亚青, 杜佳炜, 等. 不同品种高粱小曲白酒感官表征及重要风味物质对比分析[J]. 食品与发酵工业,2022,48(9):7. [SUN X X, XIONG Y Q, DU J W, et al. Comparative analysis of aroma compounds in Xiaoqu Baijiu fermented by different varieties of sorghum[J]. Food and Fermentation Industries,2022,48(9):7. [3] 程度, 曹建兰, 王珂佳, 等. 高粱对酱香型白酒品质影响的研究进展[J]. 食品科学,2022,43(7):356−364. [CHENG D, CAO J L, WANG K J, et al. Progress in understanding the effect of sorghum on the quality of Maotai-flavor Baijiu[J]. Food Science,2022,43(7):356−364. doi: 10.7506/spkx1002-6630-20210116-182 [4] 张北举, 陈松树, 李魁印, 等. 基于近红外光谱的高粱籽粒直链淀粉, 支链淀粉含量检测模型的构建与应用[J]. 中国农业科学,2022,55(1):26−35. [ZHANG B J, CHEN S S, LI K Y, et al. Construction and application of detection model for amylose and amylopectin content in sorghum grains based on near infrared spectroscopy[J]. Scientia Agricultura Sinica,2022,55(1):26−35. doi: 10.3864/j.issn.0578-1752.2022.01.003 [5] 时伟, 郑红梅, 柴丽娟, 等. 酒用高粱的营养成分及其酿造性能研究进展[J]. 食品与发酵工业, 2022, 48(21):11SHI W, ZHENG H M, CHAI L J, et al. Research progress on the nutritional components and brewing performance of brewing sorghum[J]. Food and Fermentation Industries, 2022, 48(21):11. [6] HOU Y, YUAN T J, XU J, et al. Study on identification of different producing areas of Gastrodia elata using multivariable selection and two-dimensional correlation spectroscopy of near infrared spectroscopy[J]. China Journal of Chinese Materia Medica,2019,44(4):740−749. [7] CUI J, LI P, CAO L, et al. Achievement of broadband near-infrared phosphor Ca3Y2Ge3O12: Cr3+, Ce3+ via energy transfer for food analysis[J]. Journal of Luminescence,2021,237:118170. doi: 10.1016/j.jlumin.2021.118170 [8] KHOSROSHAHI M E, PATEL Y, CHABOK R. Characterization of breast cancer antibody (anti-HER-II) conjugated on PEGylated gold nanourchin for active targeting[J]. Gold Bulletin,2022,55:149−159. doi: 10.1007/s13404-022-00316-w [9] 王海英, 杨玉珍, 任国军, 等. 利用近红外技术对河套原酒入库指标的检测研究[J]. 酿酒科技,2017(1):37−41. [WANG H Y, YANG Y Z, REN G J, et al. Rapid detection of warehousing indexes of hetao base liquor by using near infrared technology[J]. Liquor-Making Science & Technology,2017(1):37−41. doi: 10.13746/j.njkj.2016276 [10] OMAR J, SLOWIKOWSKI B, BOIX A. Chemometric approach for discriminating tobacco trademarks by near infrared spectroscopy[J]. Forensic Science International,2019,294:15−20. doi: 10.1016/j.forsciint.2018.10.016 [11] CHEN H, TAN C, LIN Z, et al. Quantifying several adulterants of Noto ginseng powder by near-infrared spectroscopy and multivariate calibration[J]. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy,2018,211:280−286. [12] SIRSOMBOON P, POSOM J. On-line measurement of activation energy of ground bamboo using near infrared spectroscopy[J]. Renewable Energy,2019,133:480−488. doi: 10.1016/j.renene.2018.10.051 [13] 王勇生, 李洁, 王博, 等. 基于近红外光谱扫描技术对高粱中粗脂肪、粗纤维、粗灰分含量的测定方法研究[J]. 中国粮油学报,2020,35(3):181−185. [WANG Y S, LI J, WANG B, et al. Research on measurement of crude fat, crude fiber and ash contents in sorghum using near-infrared reflectance spectroscopy method[J]. Journal of the Chinese Cereals and Oils Association,2020,35(3):181−185. doi: 10.3969/j.issn.1003-0174.2020.03.029 [14] 苏鹏飞, 张武岗. 基于NIR高粱淀粉含量快检技术的开发研究[J]. 酿酒科技,2022(2):107−110. [SU P F, ZHANG W G. Development of rapid detection technology for starch in sorghum based on NIR[J]. Liquor-Making Science & Technology,2022(2):107−110. [15] 苏鹏飞, 刘丽丽, 闫宗科, 等. 基于近红外高粱水分的快速分析研究[J]. 酿酒,2021,48(1):87−90. [SU P F, LIU L L, YAN Z K, et al. Study on NIR rapid analysis of water content in the sorghum[J]. Liquor Making,2021,48(1):87−90. doi: 10.3969/j.issn.1002-8110.2021.01.018 [16] PEIRIS K H S, BEAN S R, JAGADISH S V K. Extended multiplicative signal correction to improve prediction accuracy of protein content in weathered sorghum grain samples[J]. Cereal Chemistry,2020,97(5):1066−1074. doi: 10.1002/cche.10329 [17] CHADALAVADA K, ANBAZHAGAN K, NDOUR A, et al. NIR Instruments and prediction methods for rapid access to grain protein content in multiple cereals[J]. Sensors,2022,22(10):3710. doi: 10.3390/s22103710 [18] 刘敏轩, 王赟文, 韩建国. 高粱籽粒中多酚类物质的傅立叶变换近红外光谱分析[J]. 分析化学,2009,37(9):1275−1280. [LIU M X, WANG Y W, HAN J G. Determination of polyphenols in sorghum grains by near infrared reflectance spectroscopy[J]. Chinese Journal of Analytical Chemistry,2009,37(9):1275−1280. doi: 10.3321/j.issn:0253-3820.2009.09.005 [19] 王文真, 孙采云. 应用近红外反射分析高粱籽粒蛋白质, 赖氨酸和单宁含量[J]. 河南农业科学,1993(3):7−9. [WANG W Z, SUN C Y. Analysis of sorghum grain protein, lysine and tannin content using near-infrared reflection[J]. Journal of Henan Agricultural Sciences,1993(3):7−9. [20] DYKES L, HOFFMANN JR L, PORTILLO-RODRIGUEZ O, et al. Prediction of total phenols, condensed tannins, and 3-deoxyanthocyanidins in sorghum grain using near-infrared (NIR) spectroscopy[J]. Journal of Cereal Science,2014,60(1):138−142. doi: 10.1016/j.jcs.2014.02.002 [21] MABOOD F, JABEEN F, HUSSAIN J, et al. FT-NIRS coupled with chemometric methods as a rapid alternative tool for the detection & quantification of cow milk adulteration in camel milk samples[J]. Vibrational Spectroscopy,2017,92:245−250. doi: 10.1016/j.vibspec.2017.07.004 [22] YU H Y, NIU X Y, LIN H J, et al. A feasibility study on on-line determination of rice wine composition by Vis–NIR spectroscopy and least-squares support vector machines[J]. Food Chemistry,2009,113(1):291−296. doi: 10.1016/j.foodchem.2008.06.083 [23] CHEN J, LI M, PAN T, et al. Rapid and non-destructive analysis for the identification of multi-grain rice seeds with near-infrared spectroscopy[J]. Spectrochimica Acta Part A:Molecular and Biomolecular Spectroscopy,2019,219:179−185. doi: 10.1016/j.saa.2019.03.105 [24] SHI H, LEI Y, PRATES L L, et al. Evaluation of near-infrared (NIR) and Fourier transform mid-infrared (ATR-FT/MIR) spectroscopy techniques combined with chemometrics for the determination of crude protein and intestinal protein digestibility of wheat[J]. Food Chemistry,2019,272:507−513. doi: 10.1016/j.foodchem.2018.08.075 [25] GORDON R, CHAPMAN J, POWER A, et al. Unfrazzled by fizziness: Identification of beers using attenuated total reflectance mid-infrared spectroscopy and multivariate analysis[J]. Food Analytical Methods,2018,11(9):2360−2367. doi: 10.1007/s12161-018-1225-y [26] 中华人民共和国国家质量监督检验检疫总局, 中国国家标准化管理委员会. GB/T 15683-2008 大米直链淀粉含量的测定[S]. 北京: 中国标准出版社, 2008General Administration of Quality Supervision, Inspection and Quarantine of the People's Republic of China, Standardization Administration of the People's Republic of China. GB/T 15683-2008 Rice-Determination of amylose content[S]. Beijing: Standards Press of China, 2008. [27] 中华人民共和国农牧渔业部. GB 7648-1987 水稻、玉米、谷子籽粒直链淀粉测定法[S]. 北京: 中国标准出版社, 1987Ministry of Agriculture, Animal Husbandry and Fisheries of the People's Republic. GB 7648-1987 Determination of amylose of rice, maize and millet grain[S]. Beijing: Standards Press of China, 1987. [28] 中华人民共和国国家卫生和计划生育委员会, 国家食品药品监督管理总局. GB 5009.6-2016 食品安全国家标准 食品中脂肪的测定[S]. 北京: 中国标准出版社, 2016State Family Planning Commission of the People's Republic of China, State Administration for Market Regulation. GB 5009.6-2016 National Standard for Food Safety Determination of fat in food[S]. Beijing: Standards Press of China, 2016. [29] 国家市场监督管理总局, 国家标准化管理委员会. GB/T 6432-2018 饲料中粗蛋白的测定 凯氏定氮法[S]. 北京: 中国标准出版社, 2018State Administration for Market Regulation, Standardization Administration of the People's Republic of China. GB/T 6432-2018 Determination of crude protein in feed Kjeldahl nitrogen determination method[S]. Beijing: Standards Press of China, 2018. [30] 国家质量监督检验检疫总局, 国家标准化管理委员会. GB/T 15686-2008 高粱 单宁含量的测定[S]. 北京: 中国标准出版社, 2008General Administration of Quality Supervision, Inspection and Quarantine of the People's Republic of China, Standardization Administration of the People's Republic of China. GB/T 15686-2008 Determination of tannin content in sorghum[S]. Beijing: Standards Press of China, 2008. [31] SAMPAIO P S, SOARES A, CASTANHO A, et al. Optimization of rice amylose determination by NIR-spectroscopy using PLS chemometrics algorithms[J]. Food Chemistry,2018,242:196−204. doi: 10.1016/j.foodchem.2017.09.058 [32] 杨丹, 刘新, 王川丕, 等. 绿茶样品粒度对近红外光谱图和模型的影响[J]. 食品科技,2012(11):278−281. [YANG D, LIU X, WANG C P, et al. Effect of green tea particle size on near infrared spectrogram and nitrogen content model[J]. Food Science and Technology,2012(11):278−281. doi: 10.13684/j.cnki.spkj.2012.11.017 [33] BIAN X, WANG K, TAN E, et al. A selective ensemble preprocessing strategy for near-infrared spectral quantitative analysis of complex samples[J]. Chemometrics and Intelligent Laboratory Systems,2020,197:103916. doi: 10.1016/j.chemolab.2019.103916 [34] 雷敬卫, 李小庆, 白雁, 等. 近红外光谱法快速测定逍遥丸(浓缩丸)中水分含量[J]. 中国实验方剂学杂志,2013,19(19):132−135. [LEI J W, LI X Q, BAI Y, et al. Rapid determination of moisture in Xiaoyao pills (condensed) by near-infrared spectroscopy[J]. Chinese Journal of Experimental Traditional Medical Formulae,2013,19(19):132−135. [35] 余梅, 李尚科, 杨菲, 等. 基于近红外光谱技术与优化光谱预处理的陈皮产地鉴别研究[J]. 分析测试学报,2021,40(1):65−71. [YU M, LI S K, YANG F, et al. Identification on different origins of citri reticulatae pericarpium using near infrared spectroscopy combined with optimized spectral pretreatments[J]. Journal of Instrumental Analysis,2021,40(1):65−71. [36] LIU X W, CUI X Y, YU X M, et al. Understanding the thermal stability of human serum proteins with the related near-infrared spectral variables selected by Monte Carlo-uninformative variable elimination[J]. Chinese Chemical Letters,2017,28(7):1447−1452. doi: 10.1016/j.cclet.2017.03.021 [37] 熊雅婷, 李宗朋, 王健, 等. 近红外光谱波段优化在白酒酒醅成分分析中的应用[J]. 光谱学与光谱分析,2016,36(1):84−90. [XIONG Y T, LI Z P, WANG J, et al. The near infrared spectral bands optimal selection in the application of liquor fermented grains composition analysis[J]. Spectroscopy and Spectral Analysis,2016,36(1):84−90. [38] JUN L Y, CHEN H C, JIN L, et al. Near infrared determination of the content of caffeine in tea polyphenol[J]. Spectroscopy and Spectral Analysis,2005,25(8):1243−1245. [39] 李霞. 基于近红外光谱技术的赤霞珠干红葡萄酒品质指标检测方法研究[D]. 咸阳: 西北农林科技大学, 2018LI X. Rapid detection of quality index in cabernet sauvignon red wine based on near-infrared spectroscopy[D]. Xianyang: Northwest A & F University, 2018. [40] 成云玲. 基于近红外高光谱成像技术的酿酒葡萄分类及单宁含量检测[D]. 咸阳: 西北农林科技大学, 2020CHENG Y L. Detection of wine grape classification and tannins content based on near infrared hyperspectral imaging[D]. Xianyang: Northwest A & F University, 2020.