| 192 | 1 | 87 |
| 下载次数 | 被引频次 | 阅读次数 |
【背景和目的】为实现无损且准确预测烘烤过程中的烟叶含水率，探讨有效的图像特征与近红外光谱特征融合方式。【方法】首先，采集烟叶烘烤过程中的图像与光谱信息，采用过滤式方法对图像特征进行筛选，使用标准正态变量变换、基于温度项的二阶校正对原始光谱进行预处理，再利用i PLS(interval PLS)方法选取22个特征波段；接着，通过串联和外积乘法两种策略融合图像特征与近红外光谱特征，并基于递归特征消除法提取融合特征；最后，分别建立基于单一特征和融合特征的反向传播神经网络(back-propagation,BP)、支持向量回归(support vector regression,SVR)、极限学习机(extreme learning machine,ELM)、随机森林(random forest,RF)、长短时记忆神经网络(long short-term memory,LSTM)、一维卷积神经网络(one-dimensional convolutional neural network,1DCNN)预测模型。【结果】串联融合特征模型在预测准确率与精度方面比外积乘法融合特征模型、单一特征模型更胜一筹。5种串联融合特征模型测试集上的R2在0.855～0.926,RMSE在0.0690～0.0993,RPD在2.665～3.894，性能显著优于图像特征模型及近红外光谱特征模型。【结论】图像特征与近红外光谱特征的串联融合有助于提高烟叶在烘烤过程中含水率预测模型的准确性与精度，为实现烟叶含水率准确、快速且无损预测提供重要参考依据。
Abstract：[Background] Accurate and non-destructive prediction of moisture content in tobacco leaves during the curing process is essential for quality control. This study explores effective strategies for integrating image features and near-infrared(NIR) spectral features. [Methods] Image and spectral data were collected throughout the tobacco leaf curing process. Image features were selected using a filter-based method, and raw spectral data were preprocessed with standard normal variate(SNV) transformation followed by a second-order temperature-based correction. Twenty-two characteristic bands were extracted via interval partial least squares(iPLS). Two fusion strategies-concatenation and outer product multiplication-were then employed to combine image and NIR spectral features, with recursive feature elimination applied to extract fused features. Prediction models based on both single and fused features were developed using back-propagation neural networks(BP), support vector regression(SVR), extreme learning machines(ELM), random forest(RF), long short-term memory networks(LSTM), and one-dimensional convolutional neural networks(1DCNN). [Results] The concatenation fusion feature model outperformed both the outer product multiplication fusion model and the single-feature models in terms of prediction accuracy and robustness. On the test set, concatenation fusion models achieved R2 values ranging from 0.855 to 0.926, RMSE values between 0.0690 and 0.0993, and RPD values from 2.665 to 3.894, significantly surpassing models based solely on image or spectral features. [Conclusion] Concatenating image and NIR spectral features significantly improves the accuracy and precision of tobacco leaf moisture content prediction during curing, providing an important reference for developing rapid, accurate, and non-destructive prediction methods.
[1]蒋笃忠，陈洪浪，何阳，等.密集烘烤关键温度点失水率控制对烤后烟叶质量的影响[J].天津农业科学，2021, 27(05):12-15.JIANG Duzhong, CHEN Honglang, HE Yang, et al. Effect of controlling water loss rate at key Temperature points of intensive curing on the quality of tobacco leaves[J]. Tianjin Agricultural Sciences, 2021, 27(5):12-15.
[2]宫长荣.烟草调制学[M].第二版.北京：中国农业出版社，2011.GONG Changrong. Tobacco curing science[M]. 2th. Beijing:China Agriculture Press, 2011.
[3]陈飞程，杨懿德，李常军，等.基于图像信息的烤烟烘烤过程中烟叶含水率预测[J].西南农业学报，2021,34(11):2378-2384.CHEN Feicheng, YANG Yide, LI Changjun, et al. Moisture content prediction of tobacco leaf in baking process based on image information[J]. Southwest China Journal of Agricultural Sciences,2021, 34(11):2378-2384.
[4] CZAJA T, ENGELSEN S. Why nothing beats NIRS technology:The green analytical choice for the future sustainable food production[J]. Spectrochimica Acta Part A:Molecular and Biomolecular Spectroscopy, 2025,325:125028.
[5]宾俊，范伟，周冀衡，等.近红外技术结合SaE-ELM用于烤烟烘烤关键参数的在线监测[J].烟草科技，2016, 49(9):50-56.BIN Jun, FAN Wei, ZHOU Jiheng, et al. On-line monitoring of key tobacco flue-curing processing parameters by combining near infrared technology with SaE-ELM[J]. Tobacco Science&Technology, 2016, 49(9):50-56.
[6]李洋，赵鸣，徐梦瑶，等.多源信息融合技术研究综述[J].智能计算机与应用，2019, 9(05):186-189.LI Yang, ZHAO Ming, XU Mengyao, et al. A survey of research on multi-source information fusion technology[J]. Intelligent Computer and Applications, 2019, 9(05):186-189.
[7] LIU Z Y, ZHANG R T, YANG C S, et al. Research on moisture content detection method during green tea processing based on machine vision and near-infrared spectroscopy technology[J].Spectrochimica Acta Part A:Molecular and Biomolecular Spectroscopy, 2022, 271:120921.
[8]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 Chunwang, LIU Zhongyuan, YANG Ming, 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.
[9]叶文超，罗水洋，李金豪，等.近红外光谱与图像融合的杂交水稻种子分类方法研究[J].光谱学与光谱分析，2023, 43(09):2935-2941.YE Wenchao, LUO Shuiyang, LI Jinhao, et al. Research on classification method of hybrid rice seeds based on the fusion of near-infrared spectra and images[J]. Spectroscopy and Spectral Analysis, 2023,43(09):2935-2941.
[10] CHEN Y, GUO M Q, CHEN K, et al. Predictive models for sensory score and physicochemical composition of Yuezhou Longjing tea using near-infrared spectroscopy and data fusion[J].Talanta, 2024, 273:125892.
[11] DING F, LI C, ZHAI W G, et al. Estimation of nitrogen content in winter wheat based on multi-source data fusion and machine learning[J]. Agriculture, 2022, 12(11):1752.
[12]莫小明，郭磊，李贺，等.基于力声信息融合感知的香梨果肉脆度评价[J].农业工程学报，2024, 40(17):314-320.MO Xiaoming, GUO Lei, LI He, et al. Instrument detection on Korla pear flesh crispness using mechanical-acoustic information fusion[J]. Transactions of the Chinese Society of Agricultural Engineering(Transactions of the CSAE), 2024, 40(17):314-320.
[13]汪阳忠，张鑫，蔡振波，等.近红外光谱融合电子鼻数据对烟叶产地判别研究[J].河南师范大学学报：自然科学版，2024,52(02):104-110.WANG Yangzhong, ZHANG Xin, CAI Zhenbo, et al.Classification of tobacco leave parts based on the fusion of near-infrared spectroscopy and heracles electronic nose data[J].Journal of Henan Normal University(Natural Science Edition),2024, 52(02):104-110.
[14]杨睿，宾俊，苏家恩，等.基于近红外光谱与图像识别技术融合的烟叶成熟度的判别[J].湖南农业大学学报：自然科学版，2021, 47(04):406-411.YANG Rui, BIN Jun, SU Jiaen, et al. Identification of tobacco leaf maturity based on the fusion of near infrared spectroscopy and image recognition[J]. Journal of Hunan Agricultural University(Natural Sciences), 2021, 47(4):406–411.
[15]张慧，张文伟，张永毅，等.基于高光谱与纹理融合的烤烟分类方法研究[J].中国烟草学报，2022, 28(3):72-80.ZHANG Hui, ZHANG Wenwei, ZHANG Yongyi, et al. Research on classification method of flue-cured tobacco based on fusion of hyperspectral and texture features[J]. Acta Tabacaria Sinica, 2022,28(3):72-80.
[16]李增盛，孟令峰，王松峰，等.基于图像处理的烟叶烘烤阶段判别模型优选[J].中国烟草学报，2022, 28(02):65-76.LI Zengsheng, MENG Lingfeng, WANG Songfeng, et al. Selection of optimum discriminant model in tobacco curing stage based on image processing[J]. Acta Tabacaria Sinica, 2022, 28(2):65-76.
[17]杜海娜，孟令峰，王松峰，等.基于机器学习的密集烘烤过程烟叶失水率预测模型对比[J].烟草科技，2022, 55(09):81-88.DU Haina, MENG Lingfeng, WANG Songfeng, et al. Machine learning-based models for predicting dehydration rate of tobacco leaf during bulk curing and comparisons thereof[J]. Tobacco Science&Technology, 2022, 55(9):81-88.
[18]刘浩，孟令峰，王松峰，等.基于机器视觉的烤烟鲜烟成熟度判别模型优选[J].中国农机化学报，2023, 44(08):118-124.LIU Hao, MENG Lingfeng, WANG Songfeng, et al. Optimization of fresh flue-cured tobacco maturity discrimination model based on machine vision[J]. Journal of Chinese Agricultural Mechanization,2023, 44(8):118-124.
[19]韦克苏，涂永高，王丰，等.基于近红外光谱的烟叶烘烤过程质体色素实时监测[J].江苏农业科学，2021, 49(16):184-188.WEI Kesu, TU Yonggao, WANG Feng, et al. Real-time quantitative analysis of chlorophyll and carotene in tobacco during bulk curing process by near-infrared spectroscopy[J]. Jiangsu Agricultural Sciences, 2021, 49(16):184-188.
[20]褚小立，袁洪福，陆婉珍.近红外分析中光谱预处理及波长选择方法进展与应用[J].化学进展，2004(04):528-542.CHU Xiaoli, YUAN Hongfu, LU Wanzhen. Progress and application of spectral data pretreatment and wavelength selection methods in nir analytical technique[J]. Progress in Chemistry,2004(04):528-542.
[21] DU Z J, TIAN W F, TILLEY M, et al. Quantitative assessment of wheat quality using near-infrared spectroscopy:A comprehensive review[J]. Comprehensive Reviews in Food Science and Food Safety, 2022, 21(3):2956-3009.
[22]孙彦华，范永涛.近红外光谱分析中温度影响的修正[J].光谱学与光谱分析，2020, 40(06):1690-1695.SUN Yanhua, FAN Yongtao. Correction of temperature influence in near infrared spectroscopy[J]. Spectroscopy and Spectral Analysis,2020, 40(06):1690-1695.
[23] YUN Y H, LI H D, DENG B C, et al. An overview of variable selection methods in multivariate analysis of near-infrared spectra[J]. Trends in Analytical Chemistry, 2019(113):102-115.
[24]李道亮，赵晔，杜壮壮.农业领域多模态融合技术方法与应用研究进展[J].农业机械学报，2025, 56(01):1-15.LI Daoliang, ZHAO Ye, DU Zhuangzhuang. Advances in multi-modal fusion techniques and applications in agricultural field[J]. Transactions of the Chinese Society for Agricultural Machinery, 2025, 56(01):1-15.
[25] LAN T M, SHUAI S, HUAN H B, et al. A rapid prediction method of moisture content for green tea fixation based on WOA-Elman[J].Foods, 2022, 11(18):2928.
[26] SHENG X F, ZAN J Z, JIANG Y W, et al. Data fusion strategy for rapid prediction of moisture content during drying of black tea based on micro-NIR spectroscopy and machine vision[J]. Optik,2023, 276:170645.
[27] DING S S, JING J J, DOU S Q, et al. Citrus canopy spad prediction under bordeaux solution coverage based on texture-and spectral-information fusion[J]. Agriculture, 2023, 13(9):1701.
[28] YU Y, TIAN H Q, ZHANG J, et al. Detection of organic acids in maize silage based on fusion of near infrared spectroscopy and computer vision[J]. Microchemical Journal, 2025, 213:113765.
[29] GAO S, XU J H. Hyperspectral image information fusion-based detection of soluble solids content in red globe grapes[J].Computers and Electronics in Agriculture, 2022, 196:106822.
[30] SUN Z Z, TIAN H, HU D, et al. Integrating deep learning and data fusion for enhanced oranges soluble solids content prediction using machine vision and Vis/NIR spectroscopy[J]. Food Chemistry,2025, 464(01):141488.
[31] WANG F X, WANG C G. Improved model for starch prediction in potato by the fusion of near-infrared spectral and textural data[J].Foods, 2022, 11(19):3133.
基本信息:
DOI：10.16472/j.chinatobacco.2025.T0126
中图分类号:TP391.41;TS44
引用信息:
[1]王志诚,王松峰,王洪梅,等.基于多源信息融合的烤烟烘烤过程中含水率预测方法研究[J].中国烟草学报,2025,31(06):59-72.DOI:10.16472/j.chinatobacco.2025.T0126.
基金信息:
中国农业科学院科技创新工程（ASTIP-TRIC03）; 中国烟草总公司科技重点项目“基于图像精准识别的烟叶智能烘烤关键技术研究与应用”（110202102007）;中国烟草总公司四川省公司科技项目“特色品种川烟200规模化定制化开发研究与应用”（SCYC202406）