关键词: 玉米, 手机图像, 色彩参数, 关联解析, 氮营养诊断

Abstract: The development of portable diagnostic methods for monitoring crop nitrogen nutrition can effectively monitor the nutritional status of maize, accurately recommend nitrogen fertilization rate, and quickly complete the nitrogen topdressing at different growth stages of drip-irrigation maize under the condition of integrated water and fertilizer. We used smartphone camera to obtain RGB images of maize under different nitrogen treatments from four-leaf stage to milkripe stage for two years, and calculated seven common color parameters. We analyzed the correlation between color parameters and plant nitrogen concentration (N_P), constructed the model of plant nitrogen concentration and color parameter of smartphone image, and evaluated the accuracy of the model by the determination coefficient (R²),root mean square error (RMSE) and mean relative error (nRMSE). The results showed that nitrogen concentration of maize increased with increasing nitrogen application rates. In the six leaf stage, plant nitrogen concentration had a significant positive correlation with the normalized blueness intensity (NBI) and a significant negative correlation with the normalized difference yellowness index (NDYI). In the tasseling stage, plant nitrogen concentration was significantly negatively correlated with the normalized greenness index (NGI), red-green-blue vegetation index (RGBVI), and NDYI, while significantly positively correlated with the NBI. The relationship between plant nitrogen concentration and NDYI had the optimal fitting effect in the six-leaf stage, with the determination coefficients being 0.796 and 0.821 and the variation coefficients being 2.14% and 2.32% in the two years. At the tasseling stage, the fitting effect of the relationship between plant nitrogen concentration and NBIwas the best, with the determination coefficients of 0.836 and 0.773 and the variation coefficients of2.49% and 2.21% in the two years. The nitrogen nutrition diagnosis model for the plants was as follows: N_P=5.38×NDYI^-1.09 (R²=0.736) in six-leaf stage and N_P=2.07×10³×NBI^3.31 (R²=0.791)in tasseling stage. The model validation results showed that R² were0.734 and 0.790, RMSE were 0.363 and 0.746, nRMSE were 1.52% and 4.01%, respectively. TheRMSE and nRMSE values were relatively small, indicating that the model prediction accuracy was relatively high. Therefore, smartphone image parameters can be used to estimate nitrogen concentration of maize plants under drip irrigation in Ningxia, providing a theoretical basis for crop nitrogen nutrition diagnosis.

Key words: maize, smartphone image, color parameter, correlation analysis, nitrogen nutrition diagnosis.