利用不同方法反演冬小麦叶面积指数研究

Application Comparison of Partial Least Squares Regression and Stepwise Regression Methods in Winter Wheat LAI Inversion

基于高光谱数据的叶面积指数监测是快速获取冬小麦叶面积指数的重要方法。为了探究回归方法和高光谱数据变换对冬小麦叶面积指数反演精度的影响,采用逐步回归和偏最小二乘回归方法,分别建立基于冬小麦拔节期冠层高光谱数据、一阶导数光谱数据、二阶导数光谱数据和对数光谱数据的叶面积指数多元线性回归模型。结果显示,导数和对数变换能够提高冬小麦LAI反演精度,以蓝紫光、绿光、红光和近红外波段建立的一阶导数光谱数据逐步回归模型最优,建立回归模型的决定系数R2为0.974,交叉验证的RMSE为0.131,可为冬小麦LAI估算的方法选择和数据处理提供依据和参考。

Using hyperspectral data to get LAI is a key method to conduct winter wheat LAI monitor. To study the influence of regression methods and hyperspetral data conversion on winter wheat LAI inversion accuracy, basing on the winter wheat canopy original reflectance, first derivative, two derivative and logarithmic hyperspectral data, the study establishes multivariate linear regression model of LAI by step- wise regression and partial least squares regression method. The results showed that hyperspetral data transfer is helpful to improve the LAI inversion accuracy, and the best model is built by stepwise regres- sion method and based on first derivative spectrum data. In addition, coefficient of determination of the model is 0.974, and the RMSE is 0. 131. Blue violet, green, red and near infrared bands are selected by the regression model. The results can provide reference for winter wheat LAI estimate in the aspect of data selection and data processing.