冬小麦SPAD值无人机可见光和多光谱植被指数结合估算

SPAD(Soil and plant analyzer development)值能够反映作物叶片叶绿素含量,是表征作物健康状态的重要指标.采用无人机搭载可见光和多光谱相机同步获取冬小麦可见光和多光谱影像,同时获取冬小麦叶片SPAD值,探究了可见光和多光谱植被指数与SPAD值的关系,将可见光植被指数与多光谱植被指数相...

Combining UAV Visible Light and Multispectral Vegetation Indices for Estimating SPAD Value of Winter Wheat

Soil and plant analyzer development (SPAD) can reflect the chlorophyll content of crop leaves, and it is an important indicator of crop health. The visible light and multispectral images of winter wheat were synchronously obtained by unmanned aerial vehicle (UAV) equipped with visible light and multispectral cameras, and the SPAD value of winter wheat leaves were also obtained. The purposes were to explore the relationship between visible light vegetation indices, multispectral vegetation indices and SPAD value, estimate SPAD value by combining visible light and multispectral vegetation indices, and estimate SPAD value by using stepwise regression and random forest regression. The results were compared to select the best model for estimating SPAD value of winter wheat leaves. The results showed that SPAD value had good correlation with visible light vegetation indices (IKAW and RBRI) and multispectral vegetation indices (GNDVI, CI, GMSR and GOSAVI). Besides, SPAD value had a good correlation with the combination index of visible light vegetation index (CIVE) and multispectral vegetation index (GNDVI). The R2 of the estimation model of combination index was 0.89, the RMSE was 2.55, and nRMSE of the model verification was 6.21%, respectively. The results showed that compared with the indices of visible light vegetation and that of multispectral vegetation respectively, the model of stepwise regression and random forest regression of the combination indices of visible light vegetation and multispectral vegetation were more accurate in estimating SPAD value. The R2 of the optimal stepwise regression model of combination indices was 0.91, and the R2, RMSE and nRMSE of the model verification were 0.89, 2.32 and 5.64%, respectively. The R2 of the random forest regression model of combination indices was 0.90, and the R2, RMSE and nRMSE of the model verification were 0.88, 2.51 and 6.12%, respectively, which indicated good estimation results. The research result provided a reference for the estimation of winter wheat growth information based on the combining UAV visible light and multispectral image vegetation indices and improved the accuracy and stability of the estimation model.