冬小麦不同叶位叶片的叶绿素含量高光谱估算模型  

Hyperspectral Estimation Model of Chlorophyll Content in Different Leaf Positions of Winter Wheat

作　　者：马春艳[1] 王艺琳 翟丽婷 郭辅臣 李长春[1] 牛海鹏[1] MA Chunyan;WANG Yilin;ZHAI Liting;GUO Fuchen;LI Changchun;NIU Haipeng(School of Surveying and Land Information Engineering, Henan Polytechnic University, Jiaozuo 454000, China)

机构地区：[1]河南理工大学测绘与国土信息工程学院,焦作454000

出　　处：《农业机械学报》2022年第6期217-225,358,共10页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金项目(41871333);河南省高校科技创新团队支持计划项目(22IRTSTHN008)。

摘　　要：科学、高效地获取作物不同叶位叶绿素含量的垂直分布信息,可监测农作物长势状况并进行田间管理。基于冬小麦抽穗期获取的不同叶位叶片的高光谱反射率和叶绿素含量实测数据,将原始光谱、一阶微分光谱、二阶微分光谱、植被指数和连续小波系数与叶绿素含量进行相关性分析,筛选相关性较强的光谱特征参数,然后分别采用偏最小二乘回归、支持向量机、随机森林和反向传播神经网络4种机器学习算法构建冬小麦上1叶、上2叶、上3叶和上4叶的叶绿素含量估算模型,并根据精度评估结果筛选不同叶位叶绿素含量估算的最佳模型。结果表明,上1叶、上2叶和上3叶采用小波系数结合偏最小二乘回归构建的叶绿素含量估算模型精度最高,建模和验证R^(2)分别为0.82和0.75、0.80和0.77、0.71和0.62;上4叶采用植被指数结合支持向量机构建的叶绿素含量估算模型效果最佳,建模和验证R^(2)为0.74和0.79。研究结果可为基于遥感技术精准监测作物营养成分的垂直变化特征提供理论和技术支撑。The information of vertical distribution of chlorophyll content in different leaf positions of crops was obtained scientifically and efficiently to facilitate monitoring of crop growth conditions and field management.Based on the hyperspectral reflectance and chlorophyll content of different leaf positions of winter wheat obtained during the heading period,the correlation analysis of raw spectra,first-order differential spectra,second-order differential spectra,vegetation indices,continuous wavelet coefficients and chlorophyll content were performed to screen the spectral feature parameters with strong correlation.Then partial least squares regression,support vector machine,random forest and back propagation neural network algorithms were employed to construct chlorophyll content estimation models for the upper 1,upper 2,upper 3 and upper 4 leaves of winter wheat,and the best models for chlorophyll content estimation at different leaf positions were screened based on the accuracy assessment results.The results showed that the chlorophyll content estimation models constructed using wavelet coefficients combined with partial least squares were the most accurate for the upper 1,upper 2 and upper 3 leaves,with modeling and validation R^(2) of 0.82 and 0.75,0.80 and 0.77,0.71 and 0.62,respectively;the chlorophyll content estimation models constructed using vegetation indices combined with support vector machine were the best for the upper 4 leaves,with modeling and validation R^(2) of 0.74 and 0.79,respectively.The research result could provide theoretical and technical support for accurate monitoring of the vertical variation characteristics of crop nutrient content based on remote sensing technology.

关 键 词：冬小麦 不同叶位 叶绿素含量 高光谱 机器学习算法 

分 类 号：S512.11[农业科学—作物学]