Time Series Estimation Of Rice LAI Using UAV Vegetation Index And Canopy Height

Rice is very important to the food security of China and even the world.So,it is of great significance to grasp the growth status of rice efficiently,in real time and accurately.Leaf,as the main organ of photosynthesis,greatly affected the overall growth of rice.Leaf Area Index（LAI）is regarded as an important parameter to reflect the growth state of leaves.By virtue of its advantages of large area,rapid and dynamic monitoring,remote sensing method provides effective means for crop LAI monitoring,among which the empirical model constructed by vegetation index is the most widely used.However,for rice,the specific planting background and canopy structure will affect the accuracy of the vegetation index model.Therefore,it is of great practical significance to design an estimation model of rice LAI by remote sensing method which is less affected by planting background and canopy structure.In this paper,the main research object was"RL"multi-variety hybrid rice in the rice growing area of double-single-cropping rice in central China.The multi-spectral remote sensing and RGB image data obtained from the UAV platform were used to explore the variation rules of rice canopy reflection spectrum,canopy height and LAI with rice growth.Using NDVI,EVI,WDRVI,CI_green,CI_rededge,NDRE,MTCI and OSAVI to construct a multi-time sequence rice LAI estimation model based on vegetation index,canopy height and the combination of vegetation index&canopy height.After evaluating the accuracy of the above models,the estimation effects of different models are compared,and the potential factors affecting the effect and accuracy of the estimation model are discussed.The following conclusions are drawn:（1）In the rice canopy spectrum obtained by the UAV platform,the reflectivity of green band,red band,red edge band and near-infrared band are strongly correlated,while the reflectivity of red edge band and other bands is weak.With the growth of rice,the reflectance of the near-infrared band of the canopy spectrum showed a pattern of increasing then decreasing,and the reflectance of the red-edge band gradually increased.The reflectance of red edge band of different rice varieties varied greatly after heading.（2）It is necessary to distinguish the rice growth period to estimate rice LAI when using the 8 planting indexes studied in this paper to estimate rice LAI.When the rice growth period is not distinguished,the LAI estimation effect of vegetation index model is not good,and the optimal model is EVI model which R²=0.405,RMSE=1.209.When the rice growth period is distinguished,LAI estimation effect was good in the stage without heading,and the optimal OSAVI model R²=0.621,RMSE=0.985.But the modeling effect of heading stage is not good,the optimal model R²=0.1031,RMSE=0.964.（3）The full-heading stage of rice was highly correlated with the moment when the canopy height reached the maximum,with a correlation coefficient of 0.94.For the rice LAI estimation model based on canopy height,when the growth stage was not distinguished,the model coefficient of determination R²=0.629,RMSE=0.955.If the growth stage is distinguished,the model R² in the stage before heading is increased to0.702,RMSE is reduced to 0.872.And the model fitting effect and precision in the heading stage are significantly reduced,which is not as good as the vegetation index inversion LAI model in the heading stage.（4）According to the classical hypothesis on the spatial distribution of vegetation leaves in the average canopy projection theory,this paper presents a rice LAI estimation model based on combination of vegetation index&canopy height.After adding canopy height,the fitting effect and precision of various vegetation index models were improved.In the H*ln(OSA（1+1）model,R² can reach 0.66 and RMSE can be reduced to 0.915 without distinguishing the growth stage,better than the rice LAI estimation model constructed by vegetation index or canopy height alone.