Study On Hyperspectral Estimation Of Nitrogen In Corn Canopy

The No.1 Document of the Central Committee of China has repeatedly proposed to vigorously promote a new planting structure that combines farming and farming.Heilongjiang is located in the eco-pastoral zone of the north,and the new planting methods of agriculture and animal husbandry are worthy of vigorous promotion.Therefore,this paper pro poses a new planting method of agriculture and animal husbandry integration.Because the traditional nutrient measurement method requires a lot of manpower and time,it is not suitable for the nutrient measurement demand of the new farming and animal husbandry integration method,and the hyperspectral technology can realize the rapid and non-destructive speed detection of the corn canopy nitrogen.In this paper,the production model of cornfield goose in Agriculture and Animal Husbandry Integration is used as a test.The De Meyer No.1 corn canopy is used as the test object,and the nitrogen content of corn canopy is estimated by hyperspec tral detection.Research and proposed a combination of multiple algorithms to estimate the nitrogen content of corn canopy.The experiment was conducted in Yanjun Farm,Hegang City,Heilongjiang Province in the fall of 2017.The experimental field was a corn field raising geese.Firstly,the Headwall hyperspectral imaging system was used to image the corn canopy leaves of different periods and the corresponding corn canopy leaves were numbered and saved.The extracted hyperspectral blades were subjected to reflectance extraction using ENVI 5.0,and the collected corn leaves were accurately measured for total nitrogen content using an AA3 flow analyzer.Then,Savitzky-Golay smoothing(SG),standard normal transformation(SNV),multiple scattering correction(MSC),first derivative(FD),second order are used for the extracted reflectance.Five pretreatment methods of second derivative(SD),combined into 11 pretreatment methods(SG,SNV,MSC,FD,SD,SG-FD,SG-SD,SNV-FD,SNV-SD,MSC-FD And MSC-SD)respectively denoised the corn canopy in the big bell stage,grazing stage and grazing stage,and established the PLSR model with the processed data and nitrogen,and analyzed the parameters of the model to determine 11 kinds.The pretreatment method is excellent.On the basis of preprocessing,the application of the feature band algorithm is carried out,and the large projection speaker is selected by successive projections algorithm(SPA),principal component analysis(PCA)and competitive adaptive weighted sampling method.The characteristic bands of the oral period,the pre-grazing period,the late grazing period and the whole growth period are extracted from the characteristic bands of different periods.Finally,partial least squares regression(PLSR),radial bas is function neural network(RBFNN),extreme learning machine(ELM)and support vector machines are used.,SVM)to establish an estimation model for four growth periods,such as large flare period,pre-grazing period,grazing late stage and whole growth period,selected according to model determination coefficient(R2)and root mean square error(RESM)for corn field The best feature band and modeling method for goose testing.This paper proposes a multi-processing method combination model SNV-FD-CARS-SVM suitable for estimating the nitrogen content of corn canopy in agro-pastoral integration.Among them,SNV can effectively eliminate the spectral noise caused by light scattering,FD can eliminate the interference effect generated between the bands;CARS can clear a large number of repeated spectral information to select characteristic bands,in the big bell stage,the grazing stage,the grazing stage and 46,42,24,and 72 characteristic bands were extracted during the entire growth period.Comparing the other model parameters,the CARS-SVM model showed the best model in each period and was 0.9451 and 0.8952 in the big bell stage,0.9478 and 0.9002 in the early grazing period,0.9386 and 0.9021 in the late grazing period,and the whole growth period was 0.9418 and 0.9095;and 0.1114 and 0.2023 in the big bell period,0.1355 and 0.2074 in the early grazing period,0.1980 and 0.1713 in the late grazing period,and the whole growth period is 0.1455 and 0.1332.This study can provide a reference for the rapid dete ction of nitrogen content in corn canopy,and provide a theoretical basis for precise nitrogen control and dynamic balance.