Nitrogen Nutrition Diagnosis Model Of Winter Wheat Canopy Digital Image Based On Mobile Phone Image

In order to realize the rational utilization of nitrogen(N)fertilizer resources,simple and accurate diagnosis of the N nutrition status of crops is essential.Based on the characteristics of N nutrition,this research conducted experiments on N levels of different wheat cultivars in Yuzhou City,Henan Province in 2018-2019 and 2020-2021.The mobile phone camera was used to obtain the winter wheat canopy images at the reviving stage and the jointing stage of winter wheat.Using the method of color parameter average and color parameter histogram,the application of different methods in different cultivars and color parameters was analyzed;explore changes in nitrogen accumulation,plant nitrogen concentration and yield of different cultivars under various N treatments;The mean value model and histogram model of canopy image color parameters were established respectively,and the evaluation indexes and stability of different models were compared.Finally,a N nutrition diagnosis model based on the winter wheat canopy image obtained by the mobile phone camera was established.The main findings are as follows:(1)Different N application rates significantly affect the growth and development of winter wheat.From 2017-2018,2019-2020,the effects of different N fertilizer rates on plant nitrogen concentration,N accumulation and yield of different cultivars of wheat can be seen.Reasonable application of N fertilizer under different cultivars can significantly increase the N concentration of leaves and plants,and promote crop growth and development.Increase plant N accumulation,and yield usually reaches the maximum when the N application rate is180-240 kg·ha-1.The results of the two-year study showed that with the increase in the amount of N fertilizer,the plant N concentration and N accumulation continued to increase,reaching the maximum in the N360 treatment;there are differences in wheat output between years and cultivars.Among the six cultivars in 2017-2018 and 2019-2020,the output of ‘Huayu 198’and ‘Jimai 325’ reached the highest respectively.(2)The calculation method of the difference index of the histogram area of the wheat canopy image is studied.The histogram of the canopy image is obtained after ENVI processing.Based on the double peaks of the histogram in the green period,the calculation of the difference between the vegetation peak area and the soil peak area is proposed.The method has achieved good results.(3)Extract the mean value of the color parameter of the winter wheat canopy image and the area difference information of the color parameter histogram.Extract the G/R,G/B,red light standard value R/(R+G+B),green light standard value G/(R+G+B),blue light standard value B/(R+G+B)of the winter wheat canopy image and(G-R)/(R+G+B)six color parameter characteristic values,histogram sensitivity interval and area difference index.Analyzing the relationship between six color parameters and wheat N concentration,N accumulation and yield,it is found that the color parameter histogram has the best correlation with wheat N concentration and N accumulation;relative to the average value of the canopy image color parameter,the color parameter(G-R)/(R+G+B)histogram can fully reflect the crop leaf growth and the overall color information of the vegetation,and weaken the differences between different cultivars,and also have a relatively stable correlation in different growth periods.(4)A total of three strategies are used to build a model.Strategy one,color parameter mean value model;strategy two,color parameter histogram model;strategy three,neural network model.In strategy 1,the mean value model of color parameters was used to estimate the plant N concentration and N accumulation of winter wheat.In the reviving stage,the determination coefficient R2 of color parameter G/B and plant N concentration was 0.586,the mean absolute percentage error MAPE was 16.09%,and the standard error RMSE was 0.551.At the jointing stage,the R2 of NGI and RMSE were 0.504,222.13,and the MAPE was79.12%.The color parameter histogram model of strategy 2 includes histogram area difference and histogram sensitive interval model.Multiple regression method was used to establish a mathematical model of color parameter(G-R)/(R+G+B)histogram area difference index,wave peak height and nitrogen application rate to calculate the plant N concentration in the reviving stage of winter wheat.The determination coefficient was R2=0.392.After the model verification,the RMSE and MAPE of the estimated plant N concentration in the reviving stage and the actual plant N concentration were 0.499 and 13.97% respectively.In the histogram sensitive interval model,the R2 of plant N concentration model at jointing stage was 0.445,RMSE and MAPE were 0.577 and 13.04%,respectively.The R2 of N accumulation was 0.506,while RMSE was 259.88 and MAPE was 70.29%.In strategy 3,the color parameter(G-R)/(R+G+B)histogram area difference index was used to estimate the plant N concentration at joining stage by using the neural network model,and the determination coefficient was R2=0.596.The model evaluation indexes RMSE and MAPE of the model validation set were 0.448 and 12.46%,respectively.The determination coefficient of plant N accumulation at jointing stage was R2=0.819.The RMSE and MAPE of estimated and actual plant N accumulation at jointing stage were 154.88 and 30.94%,respectively,according to the model verification.In conclusion,the neural network model established by comparing the estimation effect of different models and combining with the difference index of color parameter histogram of winter wheat canopy image has a good application effect,which improves the estimation accuracy and reduces the error at the same time,and has a strong application potential in the diagnosis technology of winter wheat N nutrition.