基于冠层光谱特征和株高的马铃薯植株氮含量估算  

Estimation of Potato Plant Nitrogen Content Based on Canopy Spectral Characteristics and Plant Height

作　　者：樊意广 冯海宽[1,3] 刘杨 边明博 孟炀 杨贵军 FAN Yiguang;FENG Haikuan;LIU Yang;BIAN Mingbo;MENG Yang;YANG Guijun(Key Laboratory of Quantitative Remote Sensing in Agriculture,Ministry of Agriculture and Rural Affairs,Information Technology Research Center,Beijing Academy of Agriculture and Forestry SciencesꎬBeijing 100097,China;National Engineering Research Center for Information Technology in Agriculture,Beijing 100097,China;National Engineering and Technology Center for Information Agriculture,Nanjing Agricultural University,Nanjing 210095,China;Key Laboratory of Modern Precision Agriculture System Integration Research,Ministry of Education,China Agricultural University,Beijing 100083,China)

机构地区：[1]北京市农林科学院信息技术研究中心农业农村部农业遥感机理与定量遥感重点实验室,北京100097 [2]国家农业信息化工程技术研究中心,北京100097 [3]南京农业大学国家信息农业工程技术中心,南京210095 [4]中国农业大学现代精细农业系统集成研究教育部重点实验室,北京100083

出　　处：《农业机械学报》2022年第6期202-208,294,共8页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金项目(41601346);广东省重点领域研发计划项目(2019B020216001);2022年度农业农村部农业遥感机理与定量遥感重点实验室建设项目(PT202224)。

摘　　要：为及时准确地掌握作物的植株氮含量(PNC)信息,监测作物生长状况,实现农田氮素施肥的科学管理,以马铃薯为研究对象,首先获取了现蕾期、块茎形成期、块茎增长期、淀粉积累期和成熟期的数码影像,并实测了各生育期的PNC、株高(H)和地面控制点(GCP)的三维坐标。其次利用各生育期的无人机数码影像与GCP结合生成试验区域的数字正射影像(DOM)和数字表面模型(DSM),并从中提取冠层光谱特征和株高(H_(dsm))。然后将各生育期提取的H_(dsm)和数码影像变量与地面实测的PNC进行相关性分析,从中筛选出相关性较好的影像变量和H_(dsm)作为马铃薯PNC估算模型的输入参数。最后分别基于影像变量和影像变量结合H_(dsm)利用多元线性回归(MLR)、误差反向传播(BP)神经网络和Lasso回归3种方法构建马铃薯PNC估算模型。结果表明:基于DSM提取的H_(dsm)与实测H具有较高的拟合度(R^(2)为0.860,RMSE为2.663 cm,NRMSE为10.234%);各生育期加入H_(dsm),均能提高马铃薯PNC的估算精度和稳定性;各生育期利用MLR方法构建的PNC估算模型优于BP神经网络和Lasso回归。该研究可为马铃薯PNC状况的高效、无损监测提供技术支撑。Timely and accurate grasp of crop plant nitrogen content(PNC)information is helpful to monitor crop growth and realize the scientific management of farmland nitrogen fertilization.Based on this,taking unmanned aerial vehicle(UAV)as the platform to obtain digital images of potato budding,tuber formation,tuber growth,starch accumulation,and maturity period,and the PNC,plant height,and the three-dimensional coordinates of the ground control point(GCP)were measured.Secondly,the digital orthophoto map(DOM)and digital surface model(DSM)of the test area were generated by combining the digital images of UAV in each growth period with GCP.Then,the correlation analysis between the H_(dsm) and the constructed image variables of each growth period with the PNC measured on the ground were carried out,and the image variables with good correlation were selected as the input parameter of the potato PNC estimation models with the H_(dsm).Finally,based on the image variables and image variables combined with H_(dsm),three methods of multiple linear regression(MLR),error back propagation(BP)neural network,and Lasso regression were used to construct the PNC estimation models of potato at each growth stage.The results showed that the H_(dsm) extracted based on DSM had a high degree of fit with the measured H(R^(2) was 0.860,RMSE was 2.663 cm,and NRMSE was 10.234%).Adding H_(dsm) in each growth period can improve the accuracy and stability of estimating potato PNC.The effect of PNC estimation model constructed by MLR method in each growth period was better than that of BP neural network and Lasso regression.Therefore,the research result can provide a technical reference for the efficient and non-destructive monitoring of potato PNC status.

关 键 词：马铃薯 植株氮含量 株高 无人机 数码影像 冠层光谱特征 

分 类 号：S25[农业科学—农业机械化工程]