基于无人机多光谱的棉花叶片氮含量的估测研究
首发时间:2024-04-22
摘要:传统的叶片氮素含量测定需要破坏性取样,费时费力,估计结果也较真实值偏差较大。遥感手段以其大面积监测的优点为作物叶片含氮量监测提供了有效方法。基于常用于作物叶片氮含量估算的多光谱植被指数,分析了植被指数与棉花冠层叶片氮含量的相关性,其中NDVI、CIre、NDVIre三个植被指数与棉花冠层叶片氮素含量的相关系数绝对值较高,分别达到了0.69、0.71、0.70。将植被指数分别输入到随机森林(RFR)、XGBoost和CatBoost模型中,对比了三种模型在不同指标输入时的估测精度,对CatBoost模型进行了粒子群优化算法(PSO)改进,得到了改进的PSO-CatBoost模型以适用于棉花冠层叶片氮含量的研究,发现PSO-CatBoost模型在棉花层叶片氮含量估算中精度最高,R2为0.83,RMSE为3.42。
For information in English, please click here
Estimation of nitrogen content in cotton leaves based on UAV multispectral study
Abstract:The traditional determination of leaf nitrogen content requires destructive sampling, which is time-consuming and labour-intensive, and the estimated results also deviate from the true value. Remote sensing provides an effective method for crop leaf nitrogen content monitoring with its advantages of large area monitoring. Based on the multispectral vegetation indices commonly used for crop leaf nitrogen content estimation, the correlation between vegetation indices and cotton canopy leaf nitrogen content was analysed, in which the absolute values of the correlation coefficients of three vegetation indices, namely NDVI, CIre, and NDVIre, and the nitrogen content of cotton canopy leaves were higher, reaching 0.69, 0.71, and 0.70, respectively.The vegetation indices were inputted into the random forests (RFR), XGBoost, and Catalogues respectively. , XGBoost and CatBoost models, compared the estimation accuracy of the three models at different index inputs, and improved the CatBoost model with particle swarm optimisation algorithm (PSO), and obtained the improved PSO-CatBoost model to be applied to the study of nitrogen content of cotton canopy leaves, and it was found that the PSO-CatBoost model has the The PSO-CatBoost model was found to have the highest accuracy in the estimation of nitrogen content in cotton layer leaves, with R2 of 0.83 and RMSE of 3.42.
Keywords: words: UAV Cotton Leaf nitrogen content CatBoost
引用
No.****
动态公开评议
共计0人参与
勘误表
基于无人机多光谱的棉花叶片氮含量的估测研究
评论
全部评论0/1000