Abstract: The experiment on nitrogen fertilizer gradient of maize under drip irrigation in Ningxia Pingjibao farm was conducted from 2018 and 2019. The variety of Tianci 19 was used as an experimental material, six nitrogen level treatments were set, which was 0(N0), 90(N1), 180(N2), 270(N3), 360(N4) and 450(N5)kg(N)·ha⁻¹, respectively. P and K fertilizers was conventional fertilization treatment, which was 138kg·ha⁻¹ and 120kg·ha⁻¹. Using drip irrigation water and fertilizer integration technology, fertilizer was applied along with water. The proportion of fertilizer application in each growth stage was 10% in seedling stage, 45% in six leaves stage, 20% in silking stage and 25% in filling stage. Maize canopy images were obtained by UAV equipped with digital camera in the maize six leaves stage(V6), ten leaves stage(V10), twelve leaves stage(V12), silking stage(R1) and milk stage(R3), and used Matlab code developed by digital image recognition system to extract the corn canopy image the red light value R, green light value G and blue light value B. The correlation between 10 canopy image parameters and nitrogen nutrition indexes based on this calculation was studied, and the image color parameters with good stability and high sensitivity were screened out. The relationship model between diagnostic indices of maize nitrogen nutrition and image parameters was constructed and verified. In order to explore the feasibility of using UAV image to estimate the dynamic nitrogen nutrition of maize under drip irrigation in Ningxia Yellow River Irrigation area during the period of joining and milk ripening. The results showed that the ratio of green light to red light (G/R), standardized value of green light (NGI), and red-green-blue vegetation index (RGBVI) were highly correlated with plant nitrogen content and leaf nitrogen content and had small coefficient of variation, which could be used as the potential optimal color parameters for nitrogen nutrition diagnosis. The regression model of optimal color parameters with nitrogen content in plant and leaf was constructed, and the power function model could better predict nitrogen status of maize. The same nitrogen experiment in 2019 was used to verify the model. It was found that the R2 of measured and estimated value of NGI and plant nitrogen concentration and leaf nitrogen concentration were 0.738 and 0.689, respectively. The test indexes RMSE were 2.594 and 3.014, and nRMSE were 13.125% and 13.347%. The prediction accuracy and accuracy are higher than G/R and RGBVI. Therefore, NGI can be selected as the optimal parameter for dynamic diagnosis of nitrogen nutrition in drip irrigation maize at the stage of V6-R3 stage, and the correlation model between parameter NGI and plant nitrogen concentration (N_P=4.967×10⁶NGI^14.26) R² was 0.707. The R² of the model (N_L=1.707×10⁶NGI^12.88) was 0.654. The results showed that the application of UAV image technology could dynamically estimate the nitrogen nutrition status of maize during the period of V6-R3 stage in Ningxia Yellow River Irrigation area, and the nitrogen nutrition diagnostic model constructed could provide a theoretical basis for the precise allocation of nitrogen fertilizer for drip irrigation maize in Ningxia Yellow River irrigation area.

Key words: Drip irrigation, Maize, Image parameters, Plant nitrogen content, Leaf nitrogen content, Dynamic diagnosis