基于高光谱的烤烟叶绿素含量估算模型研究

doi:10.13733/j.jcam.issn.2095⁃5553.2022.03.014

摘要/Abstract

摘要： 研究烤烟叶片叶绿素含量与高光谱参数的相关性，建立叶绿素含量估算模型，为构建或筛选系统的烟叶烘烤特性评价指标奠定基础。以云烟87为研究对象，测定不同成熟度水平和不同烘烤温度下，叶片叶绿素含量及400~1 000 nm光谱反射率，以烤烟叶片高光谱反射率与烤烟叶片叶绿素含量为数据源，用SPA（连续投影算法）对高光谱数据进行特征波段筛选，筛选出10个与叶绿素含量相关的特征波长作为实验样本数据，采用基于SPA算法的SPA-BP，SPA-Ridge和SPA-LR3种预测模型预测不同烘烤温度点烟叶叶片叶绿素浓度，并比较各模型的决定系数（R²），均方根误差（RMSE）以及均方误差（MRE)。3种基于SPA连续投影算法的预测模型都能较好有效预测不同烘烤温度点烟叶叶片叶绿素含量，其中SPA-BP预测模型效果最好，R²达到了0.967，RMSE为0.101，SPA-LR预测模型次之，R²达到了0.956，SPA-Ridge预测模型最低，R²达到了0.916，经验证SPA-BP预测模型的准确率为83.33%，SPA-LR预测模型的准确率为75%，SPA-Ridge预测模型的准确率为70.83%，表明BP神经网络方法的预测效果要优于线性方法，具有更好的寻优能力和预测精度，预测模型可为烟叶烘烤过程中叶绿素含量的定性研究提供理论依据。

关键词: 叶绿素, 高光谱, SPA, 特征波段, BP神经网络, 预测模型

Abstract: This paper study the correlation between chlorophyll content and hyperspectral parameters of flue⁃cured tobacco leaves and establish an estimation model of chlorophyll contentto lay a foundation for establishing or screening evaluation indexes of flue⁃cured tobacco leaf curing characteristics.Tobacco type Yunyan 87 was selected as the research object. The chlorophyll content of flue⁃cured tobacco leaves and spectral reflectance of 400-1 000 nm were measured under different maturity levels and different curing temperatures,and the characteristic bands of hyperspectral data were screened by SPA (continuous projection algorithm). Ten characteristic wavelengths related to chlorophyll content were selected as experimental sample data. Three prediction models, SPA-BP, SPA-Ridge, and SPA-Ridge, based on the SPA algorithm, were used to predict the chlorophyll concentration of tobacco leaves at different curing temperatures and the coefficient of determination (R2) of each model was compared. Root mean square error (RMSE) and mean square error (MRE).The three prediction models based on the SPA continuous projection algorithm effectively predicted the chlorophyll content of tobacco leaves at different baking temperatures, and the SPA-BP prediction model was the best 0.967, RMSE is 0.101,SPA-Ridge is the second, R2=0.956,SPA⁃Ridge model was the lowest, R2=0.916. The prediction effect of the BP neural network method was better than that of the linear method, and it had better optimization ability and prediction accuracy. The prediction model can provide a theoretical basis for qualitative research of chlorophyll content in the tobacco curing process.

Key words: chlorophyll, hyperspectral, SPA, characteristic band, BP neural network, prediction model

中图分类号:

蒋柏春, 李德仑, 韦克苏, 张富贵, 王杰, 刘红芸. 基于高光谱的烤烟叶绿素含量估算模型研究[J]. 中国农机化学报, 2022, 43(3): 104-110.

Jiang Baichun, Li Delun, Wei Kesu, Zhang Fugui, Wang Jie, Liu Hongyun.. Estimation model of chlorophyll content of flue⁃cured tobacco based on hyperspectral[J]. Journal of Chinese Agricultural Mechanization, 2022, 43(3): 104-110.

参考文献

[1] 董恒，孟庆野，王金梁，等. 一种改进的叶绿素提取植被指数[J]. 红外与毫米波学报， 2012, 31(4): 336-341.
Dong Heng, Meng Qingye, Wang Jinliang, et al. A modified vegetation index for crop canopy chlorophyll content retrieval [J]. Journal of Infrared and Millimeter Waves, 2012, 31(4): 336-341.
[2] 王强，易秋香，包安明, 等. 基于高光谱反射率的棉花冠层叶绿素密度估算[J]. 农业工程学报， 2012, 28(15): 125-132.
Wang Qiang, Yi Qiuxaing, Bao anming, et al. Estimating chlorophyll density of cotton canopy by hyperspectral reflectance [J]. Transactions of the Chinese Society of Agricultural Engineering, 2012, 28(15): 125-132.
[3] 邢雪霞，刘国顺，贾方方, 等. 烤烟叶片色素含量的高光谱预测模型研究[J]. 中国烟草学报， 2014, 20(1): 54-60.
Xing Xuexia, Liu Guoshun, Jia Fangfang, et al. Hyperspectral prediction model of flue⁃cured tobacco leaf pigment content [J]. Acta Tabacaria Sinica, 2014, 20(1): 54-60.
[4] Kaasalainen, S,Nevalainen, et al. Fast and nondestructive method for leaf level chlorophyll estimation using hyperspectral LiDAR [J]. Agricultural & Forest Meteorology,2014, 198-199: 250-258.
[5] 李敏夏，张林森，李丙智, 等. 苹果叶片高光谱特性与叶绿素含量和SPAD值的关系[J]. 西北林学院学报， 2010, 25(2): 35-39.
Li Minxia, Zhang Linsen, Li Bingzhi, et al. Relationship between spectral reflectance feature and their chlorophyll contentrations and SPAD value of apple leaves [J]. Journal of Northwest Forestry University, 2010, 25(2): 35-39.
[6] 李方舟, 冯美臣, 杨武德, 等. 水旱地冬小麦叶绿素含量高光谱监测[J]. 生态学杂志, 2013, 32(12): 3213-3218.
Li Fangzhou, Feng Meichen, Yang Wude, et al. Monitoring of winter wheat chlorophyll content in irrigated and dry lands of Shanxi Province of China based on hyperspectral remote sensing [J]. Chinese Journal of Ecology, 2013, 32(12): 3213-3218.
[7] 杨杰, 田永超, 姚霞, 等. 水稻上部叶片叶绿素含量的高光谱估算模型[J]. 生态学报, 2009, 29(12): 6561-6571.
[8] 杨峰, 范亚民, 李建龙, 等. 高光谱数据估测稻麦叶面积指数和叶绿素密度[J]. 农业工程学报, 2010, 26(2): 237-243.
Yang Feng, Fan Yaming, Li Jianlong, et al. Estimating LAI and CCD of rice and wheat using hyperspectral remote sensing data [J]. Transactions of the Chinese Society of Agricultural Engineering, 2010, 26(2): 237-243.
[9] 金霞，金志芳，孙光举, 等. 分光光度法测定海水中叶绿素含量的研究[J]. 广州化工， 2010， 38(4):132-133, 136.
Jin Xia， Jin Zhifang， Sun Guangju, et al. Detection the content of chlorophyll in seawater by spectrophotometric method [J]. Guangzhou Chemical Industry, 2010, 38(4): 132-133, 136.
[10] 黄智, 昌宏涛，陈建国，等. 烤烟K326色素含量的高光谱估测研究[J]. 作物研究, 2015, 29(5): 516-521, 541.
Huang Zhi,Chang Hongtao, Chen Jianguo, et al. Study on estimation of pigment contents in flue⁃cured tobacco variety K326 by hyperspectral parameters [J]. Crop Research, 2015, 29(5): 516-521, 541.
[11] 严衍禄，赵龙莲，韩东海. 现代近红外光谱分析的信息处理技术［Ｍ] .北京：中国轻工业出版社， 2005．
[12] 曹楠宁，王加华，李鹏飞, 等. 基于GA和SCMWPLS算法的NIR光谱信息变量提取研究[J]. 光谱学与光谱分析， 2010, 30(4): 915-919.
Cao Nanning, Wang Jiahua, Li Pengfei, et al. Study on Variable Selection of NIR spectral information based on GA and SCMWPLS [J]. Spectroscopy & Spectral Analysis, 2010, 30(4): 915-919.
[13] 刘光达，魏星，张尚，等. 基于改进遗传算法优化反向传播神经网络的癫痫发作检测方法分析[J]. 生物医学工程学杂志， 2019, 36(1): 24-32.
Liu Guangda, Wei Xing, Zhang Shang, et al. Analysis of epileptic seizure detection method based on improved genetic algorithm optimization back propagation neural network [J]. Journal of Biomedical Engineering, 2019, 36(1): 24-32.
[14] Yong Q, Jie Y. Identical matrix design of the optimal decoupling control system with kernel methods [C]. International Conference on Machine Learning & Cybernetics. IEEE, 2002.
[15] Cawley G C, Talbot N, Foxall R J, et al. Heteroscedastic kernel ridge regression [J]. Neurocomputing, 2004, 57(2): 105-124.
[16] Bai B, Tan Y, Donchyts G, et al. A simple spatio⁃temporal data fusion method based on linear regression coefficient compensation [J]. Remote Sensing, 2020, 12(23): 3900.
[17] 刘燕德，姜小刚，周衍华, 等. 基于高光谱成像技术对脐橙叶片的叶绿素、水分和氮素定量分析[J]. 中国农机化学报， 2016, 37(3): 218-224.
Liu Yande, Jiang Xiaogang, Zhou Yanhua, et al. Quantitative analysis of chlorophyll,water and nitrogen for navel orange leaf based on hyper⁃spectral imaging technology [J]. Journal of Chinese Agricultural Mechanization, 2016, 37(3): 218-224.
[18] 孙焱鑫，王纪华，李保国, 等. 基于BP和GRNN神经网络的冬小麦冠层叶绿素高光谱反演建模研究[J]. 遥感技术与应用， 2007(4): 492-496.
Sun Yanxing, Wang Jihua, Li Baoguo, et al. Contrastive analysis based on neural network of winter wheat's chlorophyll concentration inversion with hyperspectral data [J]. Remote Sensing Technology & Application, 2007, 22(4):492-496.
[19] 葛春靖，张淑娟，孙海霞. 基于GA-BP神经网络玉露香梨可溶性固形物高光谱技术检测[J]. 现代食品科技, 2021, 37(5): 296-302, 278.
Ge Chunjing, Zhang Shujuan, Sun Haixia. Hyperspectral technology for determining the soluble solids content of“Yuluxiang” pear based on GA⁃BP neural network [J]. Modern food science and technology, 2021, 37(5): 296-302, 278.
[20] 赵红宁, 刘喜乐, 曹庆玺, 等. 近红外光谱法快速测定砂仁中的水分[J]. 中国实验方剂学杂志, 2016, 22(15): 33-36.
Zhao Hongning, Liu Xile, Cao Qingxi, et al. Rapid determination of moisture content in amomifructus by near infrared spectroscopy [J]. Chinese Journal of Experimental Traditional Medical Formulae, 2016, 22(15): 33-36.