木薯叶面积预测模型研究

Cassava Leaf Area Forecast Model

为探讨木薯叶面积快速准确的非破坏性测定方法,以E25、华南5号、华南6号、华南7号、华南8号、华南9号、华南101、华南124、华南8013共9个木薯品种为材料,计测叶片的裂叶数、叶长、叶宽、裂叶长、裂叶宽等叶片特征指标和叶片面积。分析叶片特征指标之间以及叶片特征指标和叶面积之间的关系,确定对叶面积影响较大的裂叶数(N)、最大裂叶长(L)、最大裂叶宽(W)和叶宽系数ρ(L/W)4个变量。通过模型拟合和优选,得到预测木薯叶面积的数学模型S=1.078L·WN+3.931ρ-17.78。模型拟合结果,决定系数R2为0.966 3、均方根误差RMSE为9.846 9。模型外部验证结果,相关系数r为0.982 5、均方根误差RMSE为9.389 9。在模型中导入反映缺刻的特征变量,可以提高叶面积预测的准确度,增加了模型的适应性。该模型预测准确度高,观测方法简单,应用方便。

In order to investigate the non-destructive determination of cassava leaf area quickly and accurately, nine cassava varieties including E25, SC5, SC6, SC7, SC8, SC9, SCIO1, SC124 and SC8013 were used to measure crack leaf number, leaf length, leaf width, crack leaf length, crack leaf width and leaf area. The characteristics of blades and the relationship between blade features and the leaf area were studied and crack leaf number （N）, the maximum crack length （L）, the maximum crack width （W） and the coefficient of crack width p（L/W） were determined to be the four variable indicators. Through the model fitting and optimization, a mathematical model to predict leaf area was built. Model fitting results revealed the coefficient of determination R2 was 0.966 3, RMSE was 9.846 9. The external validation of the model revealed the correlation coefficient r was 0.982 5, RMSE was 9.389 9. Introduced the feature variable of crack leaf in the model could improve the accuracy of prediction of the leaf area and increase the adaptability of the model. This model had a high prediction accuracy and was simple, which could be applied to scientific research and production.