文献类型：期刊文献

中文题名：基于无人机遥感的湿地松花青素含量预测

英文题名：Estimation of Anthocyanin Content in Pinus elliottii Based on UAV Remote Sensing

作者：陶学雨[1,2] 李彦杰[1] 栾启福[1] 姜景民[1]

第一作者：陶学雨

机构：[1]中国林业科学研究院亚热带林业研究所,浙江杭州311400;[2]南京林业大学林学院,江苏南京210037

年份：2021

卷号：43

期号：5

起止页码：1065-1077

中文期刊名：江西农业大学学报

外文期刊名：Acta Agriculturae Universitatis Jiangxiensis

收录：CSTPCD;;北大核心:【北大核心2020】；CSCD:【CSCD2021_2022】；

基金：国家自然科学基金青年科学基金项目(31901323);“十三五”国家重点研发计划(2017YFD0600502-2);中央级公益性科研院所基本科研业务费专项资金项目(CAFYBB2017ZA001-2-1)。

语种：中文

中文关键词：无人机;花青素;湿地松;多光谱

外文关键词：UAV;anthocyanins;Pinus elliottii;multispectrum

分类号：S771.8;S127

摘要：【目的】花青素是植物叶片中的第三大色素,功能多样,抗氧化性较强,不仅有助于叶片损伤修复,而且可以通过吸收光能而减少叶绿素对光的吸收,调节光合作用,从而起到光保护的作用。因此,了解花青素含量的动态信息,可以间接了解植物的营养健康状况,为植物培育管理提供一种可靠的参考指标。湿地松(Pinus elliottii)作为我国重要的造林树种,适时测定湿地松针叶中花青素含量对于了解湿地松生长生理状态具有重要的意义。提出一种用无人机遥感技术反演湿地松叶片花青素含量的方法。为利用多光谱无人机选择湿地松试验林中高花青素含量资源提供技术基础。【方法】以湿地松苗为试验对象,利用便携式紫外-可见光荧光仪获取湿地松冠层叶片的花青素含量实测值以及利用多光谱无人机同步获取湿地松冠层叶片的多光谱反射率。选取了5个光谱值(RED、GREEN、BLUE、NIR、REG)以及与湿地松花青素含量关系较密切的植被指数GNDVI、LCI、NDRE、NDVI、OSAVI、R/G、MACI、ARI和MARI共9种。对14种光谱指数进行不同的数据处理后分别基于偏最小二乘、支持向量机、BP神经网络方法建立花青素含量的预测模型,并分别验证,比较选出最优建模方法和预测模型。【结果】比较经5种数据预处理方法(Original(OG)、Standard Normal Variate(SNV)、blockScale(BS)、blockNorm(BN)、Detrend(DET))和5种重要变量选择方法(不处理、遗传算法与PLS回归相结合(ga_pls)、逆向变量消除(bve_pls)、正则化消除(rep_pls)、显著多元相关算法(smc))组合处理后,发现数据经过去趋势化处理和逆向变量消除处理后,基于支持向量机建立的模型为最优建模方法,此模型对湿地松冠层花青素含量预测可以达到最佳效果,验证集决定系数R2为0.61,均方根误差RMSE为1.34%。【结论】基于无人机多光谱技术,去趋势化处理和逆向变量消除处理与支持向量机相结合的建模方法可实现对湿地松冠层花青素含量的预测。
[Objective]Anthocyanin is the third major pigment in plant leaves,which has various functions and strong antioxidant capacity.It not only helps to repair the damage of leaves,but also reduces the absorption of chlorophyll to light by absorbing light energy,regulates photosynthesis and plays a role in light protection.Therefore,understanding the dynamic information of anthocyanin content can indirectly understand the nutritional and health status of plants,and provide a reliable reference index for plant cultivation and management.Pinus elliottii is an important afforestation tree in China.Timely determination of anthocyanin content in Pinus elliottii needles is of great significance for understanding the physiological state of Pinus elliottii growth.A method of retrieving anthocyanin content in leaves of Pinus elliottii by unmanned aerial vehicle(UAV)remote sensing technology is proposed.It provides a technical basis for the selection of high anthocyanin content resources in Pinus elliottii forest by using multispectral UAV.[Methods]Taking Pinus elliottii seedlings as the experimental object,the measured values of anthocyanin content in canopy leaves of Pinus elliottii were obtained by portable ultraviolet visible fluorescence instrument(MULTIPLEX RESEARCH),and the multispectral reflectance of canopy leaves of Pinus elliottii was simultaneously obtained by multispectral UAV.Five spectral values(RED,GREEN,BLUE,NIR,REG)and nine vegetation indices which are closely related to the anthocyanin content of Pinus elliottii(Green normalized vegetation index(GNDVI),Leaf chlorophyll index(LCI),Normalized difference rededge index(NDRE),Normalized difference vegetation index(NDVI),Optimize soil regulation vegetation index(OSAVI),Red/green index(R/G),Modified anthocyanin content index(MACI),Anthocyanin reflectance index(ARI)and Modified anthocyanin reflectance index(MARI))were selected.After different data processing of 14 kinds of spectral indices,the prediction models of anthocyanin content were established based on partial least squares,support vector machine and BP neural network,and verified respectively.The optimal modeling method and prediction model were selected by comparison.[Results]By comparing the coefficient of determination(R2)and root mean square error(RMSE)after the combination of five data preprocessing methods(original(OG),standard normal variable(SNV),blockscale(BS),blocknormal(BN),detrend(DET))and five important variable selection methods(no processing,ga_pls,bve_pls,rep_pls,smc),it was found that the model based on support vector machine was the optimal modeling method after data detrend processing and backward variable elimination.This model could achieve the best performance in predicting the content of anthocyanin in the canopy of Pinus elliottii.The decision coefficient(R^(2))of validation set was 0.61 and the root mean square error(RMSE)was 1.34%.[Conclusion]Based on UAV multispectral technology,the modeling method of combining detrended processing and reverse variable elimination processing with support vector machine is the optimal modeling method,which can predict the anthocyanin content of Pinus elliottii canopy.At the same time,the feasibility of using multi spectral UAV to select high anthocyanin content resources in Pinus elliottii experimental forest has been proved.