基于分数阶微分的春小麦叶绿素含量高光谱估算  

Spectral Estimation of Chlorophyll Content in Spring Wheat Leaves Based on Fractional Differential

作　　者：茹克亚·萨吾提 阿不都艾尼·阿不里[2,3] 尼加提·卡斯木 李虎 吾木提·艾山江[1,2] 亚森江·喀哈尔 李晓航 RUKEYA Sawuti;ABUAINI Abula;NIJATI Kasimu;LI Hu;WUMUTI Aishanjiang;YASENJIANG Kahaer;LI Xiaohang(College of Resources and Environmental Science,Xinjiang University,Urumqi,Xinjiang 830046,China;Key Laboratoryof Oasis Ecology,Xinjiang University,Urumqi,Xinjiang 830046,China;Institute of Arid Ecology and Environment,Urumqi,Xinjiang 830046,China;Huizhuo University,Anhui Province High-Resolution Earth ObservationSystem Data Products and Application Software R&D Center,Chuzhou,Anhui 239000,China)

机构地区：[1]新疆大学资源与环境科学学院,新疆乌鲁木齐830046 [2]新疆大学绿洲生态教育部重点实验室,新疆乌鲁木齐830046 [3]新疆大学干旱生态环境研究所,新疆乌鲁木齐830046 [4]滁州学院,安徽省高分辨率对地观测系统数据产品与应用软件研发中心,安徽省滁州市239000

出　　处：《麦类作物学报》2019年第6期738-746,共9页Journal of Triticeae Crops

基　　金：国家自然科学基金项目(51704259);新疆维吾尔自治区自然科学基金项目(2017D01C065)

摘　　要：为提高农作物叶片叶绿素含量高光谱估算的准确度,以阜康农作物试验地为研究靶区,测定了165个采样点的春小麦叶片叶绿素含量和叶片光谱反射率,运用分数阶微分算法进行光谱预处理,最后运用偏最小二乘法(PLSR)建立叶绿素含量估算模型。结果表明,对数学变换(√R、lgR、1/lgR、1/R)的光谱及原始光谱(R)的数据进行0~2阶分数阶微分预处理时,通过0.01水平显著性检验的波段数量明显增加,且光谱数据经4种数学变换后均在1.2阶微分与小麦叶绿素含量有较高的相关性。1.2阶微分处理后,对叶绿素含量的敏感波段数量表现为√R>lg^R>1/R>1/lg^R>R。利用对数变换和1.2阶微分计算的植被指数(NDVI、DVI、RVI、MSR705、MSR670,800、CI)建立的PLSR模型的估算精度最优,其预测的相对误差、决定系数和平方根差分别为2.17、0.87和0.243mg·g^-1,可作为小麦叶片叶绿素含量的最佳估算模型,也说明对光谱数据进行数学转换和分数阶微分处理可显著提高春小麦叶绿素含量的估算精度。To improve the estimation accuracy of prediction model of chlorophyll contents,fractional differential method was applied to deal with process the hyperspectral data of wheat leaf.The Fukang crop experimental base was selected as the study area,and chlorophyll contents and hyperspectral data of spring wheat leaf were collected from 165sampling points.The results showed that correlation between fractional differential of transformed reflectance(√R,lgR,1/lgR,1/R)and chlorophyll contents were higher than that between raw spectra and chlorophyll contents.Four kinds of mathematical transformations of spectral data showed a high correlation between 1.2-order differential and chlorophyll content in wheat.The number of sensitive bands to chlorophyll was TR>lgR>1/R>1/lgR>R. The number of sensitive bands significantly increased at 1.2-order fractional differential of transform. And 1.2-order fractional differential of transform√R had 121sensitive bands.The number of sensitive bands was decreasing the following sequence of√R >lgR>1/R>1/lgR>R.Prediction model accuracy indicated that the prediction model 2-lgR PLSR,that was built by variables of vegetable indices(NDVI, DVI,RVI,MSR705,MSR670,800and CI calculated by 1.2-order fractional differential of lgR,has a high predicting prediction accuracy with r2=0.87and relative percent deviation RPD=2.17.It could be concluded that fractional differential processing had potential in hyper-spectral estimation of chlorophyll content and large scale spatial distribution inversion.The results provide a new idea for nondestructive and accurate evaluation of chlorophyll content in winter wheat.

关 键 词：高光谱 分数阶微分 叶绿素含量 春小麦 

分 类 号：S512.1[农业科学—作物学] S314