摘要
【目的】叶绿素是植物光合作用的一个重要生理生化指标,直接反映了植物生长状况。毛竹(Phyllostachys edulis(Carriere)J.Houz.)具有重要的经济和生态效益,在固碳减排上发挥重要的作用,因此对毛竹叶绿素遥感实时动态监测研究有助于评估其生长状态和固碳能力。目前估算植物叶绿素大多数是使用实地测量法或采用卫星遥感获得植物光谱数据,建立叶绿素反演模型,但是这种反演模型时效性和准确性较弱。随着无人机技术的发展,为更精准和快速地反演叶绿素含量提供了新的方法。【方法】研究采用CCM-200手持式叶绿素仪获取毛竹相对叶绿素含量(chlorophyll content index,CCI)长时序数据,通过无人机可见光传感器获取毛竹红绿蓝(RGB)可见光影像数据,在此基础上分析基于无人机RGB影像得到的植被指数和CCI之间的相关关系,采用一元线性回归和以高斯误差函数作为隐含层激活函数的BP算法前馈神经网络(Erf-BP)两种方法建立毛竹CCI遥感反演模型并作精度评价。【结果】在构建的24个变量中,(2b-r-g)-(1.4r-b)、B/R、川岛指数(I_(KAW))、(2b-r)/3r这4个植被指数和毛竹CCI相关性较高,相关系数分别为0.927、0.916、-0.915、0.913。由(2b-r-g)-(1.4r-b)建立的一元线性模型的拟合值和验证值与实测值之间的决定系数(R^(2))分别为0.8893和0.8535,均方根误差(RMSE)分别为1.3778和1.9114。Erf-BP神经网络模型R2分别为0.8724和0.8825,RMSE分别为1.6429和1.5547。Erf-BP神经网络的验证精度要稍高于一元线性回归模型。在基于无人机RGB影像反演毛竹相对叶绿素时可选择神经网络模型。【结论】研究结合统计方法和无人机RGB影像建立了毛竹相对叶绿素反演模型,可以精准反演毛竹相对叶绿素含量,为动态监测毛竹叶绿素含量提供了可行的方法。
[Objective]An an important physiological and biochemical index of plant photosynthesis,Cholorophyll indicates the growth of the plant.Moso bamboo(Phyllostachys edulis(Carrière)J.Houz.)has important economic and ecological value,and plays an important role in carbon sequestration and emission reduction.Therefore,research on real-time dynamic remote sensing monitoring of chlorophyll of Moso bamboo can help to assess its growth state and carbon sink capacity.Currently,the plant chlorophyll is estimated mainly by using the field measurement or establishing inversion models based on the spectral data from satellite remote sensing,however,the timeliness and accuracy of the inversion models are weak.The development of UAV technology provides a new method for more accurate and rapid inversion of chlorophyll content.[Method]This study used CCM-200 handheld chlorophyll meter to obtain the long-term time series data of relative chlorophyll content index(CCI)of Moso bamboo,and the red-green-blue(RGB)image data of the bamboo were obtained by Unmanned Aerial Vehicle(UAV)visible light sensor.On this basis,the correlation between vegetation index derived from UAV RGB images and CCI was analyzed.Furthermore,the remote sensing inversion model of CCI of Moso bamboo was developed and evaluated using one-dimensional linear regression and backpropagation neural network with Gaussian error function as the activation function of the hidden layer(Erf BP).[Result]The results showed that,among the 24 variables,(2b-r-g)-(1.4r-b),B/R,Kawashima index(I_(kaw)),and(2b-r)/3r had high correlation with CCI of Moso bamboo,with correlation coefficients of 0.927,0.916,-0.915,and 0.913,respectively.The coefficient of determination(R^(2))between the fitted values and the measured values as well as that between the validated vaules and the measured values were 0.8893 and 0.8535 respectively in the one-dimensional linear model driven by the(2b-r-g)-(1.4r-b).The root mean square errors(RMSE)for the fitted and validated values were 1.3778 and 1.9114 in the one-dimensional linear model driven by the(2b-r-g)-(1.4r-b).The R2 between the fitted values and measured values as well as that between the fitted values and the measured values were 0.8724 and 0.8825 respectively for the Erf-BP neural network model.The RMSE for the fitted and validated values were 1.6429 and 1.5547 respectively in the Erf-BP neural network model.The accuracy of the Erf-BP neural network model was slightly higher than that of the one-dimensional linear model for the validated dataset.Therefore,it is better to choose the Erf-BP neural network model for estimating the CCI of Moso bamboo.[Conclusion]In this study,statistical methods and UAV visible light images were combined to develop an inversion model for estimating chlorophyll of Moso bamboo,which can accurately inverse the relative chlorophyll content and provides a feasible method for dynamic monitoring of the content of Moso bamboo.
作者
曲怡铃
唐燕
周忠生
颜芳
王聚中
徐怀兴
胡璐璐
徐小军
QU Yiling;TANG Yan;ZHOU Zhongsheng;YAN Fang;WANG Juzhong;XU Huaixing;HU Lulu;XU Xiaojun(State Key Laboratory of Subtropical Silviculture/Key Laboratory of Carbon Cycling in Forest Ecosystems and Carbon Sequestration of Zhejiang Province/School of Environmental and Resources Science,Zhejiang A&F University,Lin’an,Zhejiang 311300,China;Lin’an Agricultural and Rural Information Service Center,Lin’an,Zhejiang 311300,China)
出处
《江西农业大学学报》
CAS
CSCD
北大核心
2022年第1期139-150,共12页
Acta Agriculturae Universitatis Jiangxiensis
基金
国家自然科学基金项目(31870619)。
关键词
毛竹
叶绿素
无人机
机器学习
模型
Moso bamboo
chlorophyll
UAV
machine learning
modeling
