论文部分内容阅读
基于人工神经网络理论,针对高光谱遥感中数据冗余问题,本文建立了基于遗传算法(GA)的广义回归神经网络(GRNN)模型,利用回归分析问题中参数筛选方法,对表征冬小麦叶片全氮的光谱参数进行了筛选,并和线性回归方法对比,线性回归方法的均方根误差(RMSEP):在冬小麦叶片氮含量为34.0g kg-1~62.5g kg-1预测范围内,逐步回归模型为14.4g kg-1,后向选择为11.8g kg-1,而广义回归神经网络为3.40g kg-1。说明神经网络方法所筛选到的光谱参数更能反映小麦叶片全氮含量,且神经网络模型预测精度高。
Based on the artificial neural network theory, aiming at the problem of data redundancy in hyperspectral remote sensing, a generalized regression neural network (GRNN) model based on genetic algorithm (GA) is established. By using parameter screening method in regression analysis, (RMSEP): Compared with the linear regression method, the RMSEP of stepwise regression model was used in the prediction of leaf nitrogen content of winter wheat leaves from 34.0g kg-1 to 62.5g kg-1, Was 14.4g kg-1 and the back selection was 11.8g kg-1, whereas the generalized regression neural network was 3.40g kg-1. It shows that the spectral parameters screened by the neural network method can better reflect the total nitrogen content of wheat leaves and the prediction accuracy of the neural network model is high.