融合地面高光谱反射率和无人机RGB影像协同估测水稻氮营养指数OA

[Objective]To explore the possibility and potential of estimating the nitrogen nutrition index(NNI)of rice by combining the vegetation index(VI)calculated from the spectral reflectance of rice canopy and the color index(CI)and texture index(TI)calculated from the unmanned aerial vehicle RGB image data.[Method]This study selected two rice varieties and designed three nitrogen fertilizer gradients and two nitrogen fertilizer strategies for field experiments.RGB images and canopy spectral re-flectance were obtained during the tillering,jointing,and full-heading stages of rice.Select sensitive re-mote sensing parameters through correlation analysis,use VI,CI,TI,and their combinations during the tillering,jointing,and full-heading stages,and establish an NNI estimation model using partial least square(PLS),random forest(RF),and support vector regression(SVR)regression algorithms to verify the reliability of the estimation model.[Result]In addition to the double-peak canopy nitrogen index(DCNI),the VI,CI,and TI were significantly correlated with the NNI of rice.For PLS,RF,and SVR regression models,the combination of VI,CI and TI improved estimation accuracy compared to using VI or CI and TI alone,with the RF regression model performing the best.During the tillering,jointing,and full-heading stages of rice,the estimation accuracy of the RF model is the highest,with coefficient of de-termination(R2)values of 0.858,0.826,and 0.789,respectively and the validation R2 values are 0.622,0.605,and 0.643,respectively.[Conclusion]Therefore,the comprehensive utilization of canopy spec-tral reflectance and RGB image information has great potential to improve the accuracy of rice NNI esti-mation.This study combines ground-based hyperspectral remote sensing and unmanned aerial vehicle re-mote sensing technology,providing a new approach for rapid,non-destructive,and accurate diagnosis of nitrogen dynamics in rice.