基于特征光谱参数的叶片和冠层尺度茶多酚含量估算  

Estimation of Leaf and Canopy Scale Tea Polyphenol Content Based on Characteristic Spectral Parameters

作　　者：段丹丹 刘仲华[1] 赵春江[2,3] 赵钰 王凡 DUAN Dan-dan;LIU Zhong-hua;ZHAO Chun-jiang;ZHAO Yu;WANG Fan(College of Horticulture,Hunan Agricultural University,Changsha 410128,China;National Engineering Research Center for Information Technology in Agriculture,Beijing 100097,China;Intelligent Equipment Research Center,Beijing Academy of Agriculture and Forestry Sciences,Beijing 100097,China)

机构地区：[1]湖南农业大学园艺学院,湖南长沙410128 [2]国家农业信息化工程技术研究中心,北京100097 [3]北京市农林科学院智能装备技术研究中心,北京100097

出　　处：《光谱学与光谱分析》2024年第3期814-820,共7页Spectroscopy and Spectral Analysis

基　　金：国家自然科学基金项目(21974012);广东省科技计划项目(2019B090905006);广东省重点领域研发计划项目(2018B020241001);韶关市智慧生态茶园建设及碳储量监测技术研究项目(210909114530725)资助。

摘　　要：茶多酚具有很强的生理活性和抗氧化性,是茶品质的重要属性之一。相比传统茶多酚含量的测定方法,遥感技术监测茶多酚含量具有高效、精确及实时的优势,但如何利用遥感数据监测不同时期的茶多酚含量研究较少。该研究以广东省英德市的5个茶园的茶叶为研究对象,对春茶、夏茶和秋茶的叶片与冠层两个尺度的茶多酚含量及对应高光谱数据进行测定,利用标准正态变量变换(SNV)对叶片和冠层的高光谱反射率数据进行预处理;然后,分别采用连续投影算法(SPA)和竞争性自适应重加权采样算法(CARS)筛选不同生长季节叶片和冠层两个尺度茶多酚的敏感波段;最后,通过偏最小二乘法(PLS)、随机森林(RF)和多元线性回归(MLR)分别构建不同时期的茶多酚含量模型并进行验证。结果表明:(1)茶多酚的含量随着季节推移显著增加,春茶茶多酚含量(15.37%)最低,夏茶茶多酚含量次之(18.29%),秋茶茶多酚含量(秋茶20.77%)最高;(2)不同敏感波段筛选的茶多酚含量的光谱特征波段主要为2100~2200 nm附近、1300~1400 nm附近、红波-红边波段及绿波段;(3)基于春茶、夏茶和秋茶冠层光谱特征构建的茶多酚模型中CARS-PLS、SPA-MLR和CARS-PLS模型精度最高,建模集R2分别为0.56、0.45和0.52,RMSE分别为1.15、1.68和1.77;验证集R2分别为0.43、0.40和0.41,RMSE分别为1.60、1.91和1.91;基于春茶、夏茶和秋茶冠层叶片光谱特征构建的茶多酚模型中SPA-PLS、CARS-PLS和SPA-MLR模型精度最高,建模集R2分别为0.50、0.42和0.42,RMSE分别为1.25、1.70和1.66;验证集R2分别为0.43、0.36和0.38,RMSE分别为1.44、1.96和2.49。研究结果表明,基于遥感数据进行不同季节的叶片和冠层两个尺度的茶多酚含量估算是可行的,在大面积实时监测茶品质特征方面具有较大的潜力。The content of tea polyphenols has strong physiological activity and antioxidant properties,which are important attributes of tea quality,and play an important role in human body fat metabolism and scavenging free radicals.Compared with the assay method of tea polyphenol content,although monitoring the content of tea polyphenols based on remote sensing technology has the advantages of high efficiency,accuracy and real-time,there are few studies on how to use remote sensing data to monitor the content of tea polyphenols.This study took tea leaves from five tea gardens in Yingde City,Guangdong Province,as the research object,and measured the content of tea polyphenols and the corresponding hyperspectral data at two scales of leaf and canopy of spring tea,summer tea and autumn tea.Standard normal variate transformation(SNV)was used to preprocess leaf and canopy hyperspectral reflectance data;then,Successive projections algorithm(SPA)and competitive adaptive weighted sampling algorithm(CARS)to select the sensitive bands of tea polyphenols at two scales of leaf and canopy in different growing seasons;Finally,the tea polyphenol content models in different periods were constructed and verified by partial least squares(PLS),random forest(RF)and multiple linear regression(MLR).The results showed that:(1)The tea polyphenols content increased significantly with the passage of seasons,the content of tea polyphenols in spring tea(15.37%)was the lowest,the content of tea polyphenols in summer tea was the second(18.29%),and the content of tea polyphenols in autumn tea(20.77%in autumn tea)was the highest;(2)The characteristic bands of tea polyphenols are mainly concentrated in the short-wave near-infrared band(around 2100~2200 nm),near-infrared(around 1300~1400 nm),red wave-red edge band and green band;(3)The CARS-PLS,SPA-MLR and CARS-PLS have the highest precision among the tea polyphenol models constructed based on the spectral characteristics from spring,summer and autumn canopy,with R 2 of 0.56,0.45 and 0.52 respectively,and

关 键 词：茶多酚 高光谱 偏最小二乘法 随机森林 多元线性回归 

分 类 号：S127[农业科学—农业基础科学]