厚竹叶面积预测模型研究

Study on prediction model of leaf area for Phyllostachys edulis‘Pachyloen’

厚竹是毛竹优良品种,其叶片比毛竹小,竹叶易失水卷曲,叶面积测量误差大。基于前期毛竹叶面积预测模型研究,根据叶形对厚竹叶片进行聚类分析并将之分为3类,通过实际叶面积和叶片长宽之积进行建模,以叶面积仪所测量数据作为对照,并利用RMSE、SSE以及预测精度3个指标来验证模型预测精度。结果表明:厚竹叶片按长宽比分为3类,长宽比≤7.38、7.38<长宽比≤8.58、长宽比>8.58;按叶形的分类拟合与不分类的整体拟合R2值均大于0.95,远超叶面积仪测量结果(R2值为0.89),模型预测精度分别为95.94%、95.39%;分类拟合的均方根误差和残差平方和均最小,仅为0.40和39.88。

Bamboo leaves are easy to lose water and curl,which increase the large measurement error of leaf area.Phyllostachys edulis‘Pachyloen’is an excellent variety of Ph.edulis.Its leaves are smaller than those of Ph.edulis.Based on the previous research on the leaf area prediction model of Ph.edulis,the Ph.edulis‘Pachyloen’leaves were clustered and divided into three categories according to the leaf shapes.Modeling was carried out by the product of actual leaf area and leaf length and width,and the measured data of leaf area meter were used as a control,and the prediction accuracy of the model was verified by using RMSE,SSE and prediction accuracy.The results showed as follows:Ph.edulis‘Pachyloen’leaves were divided into three categories according to the ratio of length to width,with the ratio of length-width equal or lesser than 7.38,in the range of 7.38 to 8.58 and aspect ratio exceed 8.58;The R2 values of the whole fitting according to leaf shape classification and non-classification were both greater than 0.95,far exceeding the measurement results of leaf area meter(R2 value was 0.89),and the prediction accuracy of the model was 95.94%and 95.39%,respectively.The root mean square error and residual sum of squares of the classification fit were the smallest,only 0.40 and 39.88.