基于多层级特征筛选和无人机影像的冬小麦植株氮含量预测  

Predicting nitrogen content in winter wheat plants using multi-level sensitive feature filtering and UAV imagery

作　　者：郭燕[1,2] 王来刚 贺佳[1,2] 井宇航 宋晓宇[3] 张彦 刘婷 GUO Yan;WANG Laigang;HE Jia;JING Yuhang;SONG Xiaoyu;ZHANG Yan;LIU Ting(Institute of Agricultural Information Technology,Henan Academy of Agricultural Sciences/Key Laboratory of Huang-Huai-Hai Smart Agricultural Technology,Ministry of Agriculture and Rural Affairs,Zhengzhou 450002,China;Henan Engineering Laboratory of Crop Planting Monitoring and Warning,Zhengzhou 450002,China;Information Technology Research Center,Beijing Academy of Agriculture and Forestry Sciences,Beijing 100094,China)

机构地区：[1]河南省农业科学院农业信息技术研究所/农业农村部黄淮海智慧农业技术重点实验室,郑州450002 [2]河南省农作物种植监测与预警工程研究中心,郑州450002 [3]北京市农林科学院信息技术研究中心,北京100094

出　　处：《农业工程学报》2024年第12期174-182,共9页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家重点研发计划项目(2022YFD2001105);河南省重点研发与推广专项(232102111030);河南省农业科学院自主创新项目(2023ZC062);河南省农业科学院农业信息技术研究所科技创新领军人才培育计划项目(2022KJCX01)。

摘　　要：氮素是冬小麦生长发育必不可少的大量元素,无人机超高分辨率影像丰富的光谱信息和纹理信息为冬小麦植株氮含量精准预测提供了重要的技术途径,但是过多变量造成了信息冗余和模型复杂的问题。针对此问题,该研究提出了一种“相关分析+共线性分析+LASSO(least absolute shrinkage and selection operator)特征筛选”的多层级植株氮含量敏感特征的筛选方法,引入约束系数向量的L1正则化实现特征的稀疏性,将某些特征的系数缩小为0,基于冬小麦关键生育期(拔节期、孕穗期、开花期、灌浆期)无人机影像提取的65个光谱和纹理特征,采BP神经网络(back propagation,BP)、Adaboost、随机森林(random forest,RF)和线性回归(linear regression,LR)4种机器学习算法构建了冬小麦植株氮含量预测模型。结果表明:相关分析筛选出51个通过0.01显著性检验的变量;基于共线性分析,当LASSSO正则化参数λ取值为0.08时,17个敏感特征变量被筛选。基于筛选的敏感特征变量,BP、Adaboost、RF和LR 4种算法建立的植株氮含量预测模型均达到了0.01水平差异显著性,且BP、Adaboost和RF 3种预测模型的精度具有高度的一致性,模型R2均为0.81,RMSE分别为0.36%、0.38%和0.37%,说明该研究提出的多层级特征筛选方法不仅使得模型变得简洁,而且稳健性高,可为智慧农业氮肥精准监测、智慧管理提供技术支撑。Nitrogen is one of the most essential elements for the growth and development of winter wheat.Unmanned aerial vehicle(UAV)imagery can be expected to accurately predict the nitrogen content in winter wheat plants,due to the spectrum and texture with the ultra-high resolution.However,too many variables have resulted in the parameters redundancy and the complex structure of model.Therefore,it is very necessary to screen the model-sensitive variables for the high accuracy,in order to improve the computational efficiency of the simple model.In this study,a multi-level filtering of sensitive feature was proposed,namely,"correlation analysis+covariance analysis+LASSO(least absolute shrinkage and selection operator)feature screening",which introduced L1 regularization of the constraint coefficient vectors to achieve sparsity of the features,and reduced the coefficients of some features to zero.A total of 65 feature variables were obtained using the spectral and texture information that extracted from the UAV images of winter wheat during the critical growth stages(jointing,booting,flowering,and filling stages).65 feature variables were firstly correlated with the nitrogen content of the plant.51 feature variables were retained to pass the 0.01 level of significance.After collinearity analysis,the LASSO was introduced to filter the feature variables,in order to eliminate the collinearity among the feature variables.Back-propagation(BP)neural network,adaboost,random forest(RF)and linear regression(LR)were used to predict the nitrogen content of winter wheat plants.Once the value of LASSO regularization parameterλwas 0.08,17 sensitive feature variables were screened.The predictive models were then established for the plant nitrogen content.All BP,Adaboost,RF,and LR reached the significant differences at the 0.01 level.The accuracies of BP,Adaboost,and RF were well consistent with the model R2 of 0.81 and the root mean square errors(RMSEs)of 0.36%,0.38%,and 0.37%,respectively.The R2 and RMSE were evaluated on the training,c

关 键 词：无人机 冬小麦 氮素 多层级特征筛选 LASSO回归 机器学习 最小二乘回归 

分 类 号：S147.2[农业科学—肥料学] S24[农业科学—农业基础科学]