冬小麦生育前期LAI高光谱反演研究

Comparative studying on hyperspectral inversion of LAI in the early growth stage of winter wheat

【目的】冬小麦生育前期稀疏植被条件下叶面积指数反演对于播期、早期苗情监测有重要意义。【方法】文章利用实测冬小麦生育前期冠层高光谱数据,基于相关关系矩阵图筛选7个新的敏感植被指数、优选40个前人研究的双波段组合或多波段组合植被指数,利用单变量回归和偏最小二乘多变量回归分析47个植被指数与稀疏冬小麦叶面积指数(LAI)的相关性。【结果】植被指数PVR(650,550)、VARI(680,555,480)、RVI(1868,1946)与LAI相关性好,其中PVR(650,550)与LAI构建的模型拟合度最好,决定系数R2为0.730,均方根误差RMSE为0.450。而相对单个植被指数,利用多个植被指数的偏最小二乘多元回归模型提高了LAI估算精度,R2为0.779,RMSE为0.380。【结论】在冬小麦生育前期植被稀疏条件下,利用高光谱数据反演冬小麦LAI是可行的,可为冬小麦早期长势遥感监测提供支撑。

[Purpose]The inversion of leaf area index(LAI)under sparse vegetation condition in early growth stage of winter wheat is of great significance for the monitoring of sowing date and early seedling situation.[Method]In this paper,using the measured hyperspectral data of winter wheat canopy in the early growth stage,7 new sensitive vegetation indices were selected by correlation matrix,40 dual bands or multi bands vegetation indices were selected from the previous research,and the correlation between 47 vegetation indices and LAI of early growth stage winter wheat was analyzed by using univariate regression and partial least square multivariate regression.[Result]PVR(650,550),VARI(680,555,480),RVI(1868,1946)had a good correlation with LAI,and PVR(650,550)was highly fitted with the model constructed by LAI(the coefficient of determination R2 was 0.730,and the root mean square error RMSE was 0.450).Compared with the single vegetation index,the least squares multiple regression(PLS)model with multiple vegetation indices improved the accuracy of LAI estimation,and R2=0.779,RMSE=0.380.[Conclusion]The results show that it is feasible in inversion LAI of early growth stage of winter wheat by hyperspectral data,which can give support for wheat condition monitoring using remote sensing in wheat early stage.