Study On Method Of Beet N Diagnosis Based On Hyperspectral RemoteSensing And Digital Image Information

In this paper, study on method of beet N diagnosis based on hyperspectral remote sensing and digital image information, analysis of the research progress of crop nitrogen nutrition diagnosis at home and abroad, In this study, combined with the characteristics of the digital image and Near Infrared Spectroscopy data, Establish a Beet Nutrition Diagnosis system based on digital image and Near Infrared Spectroscopy. The main research work and achievements are as follows:1. Beet canopy VNIR is collected by ASD field spectrometer portable, digital image data is collected by Canon digital cameras,and determination to indicators of beet canopy nitrogen content,146 samples were selected as the modeling set,48 samples as the prediction set, used to establish the diagnostic model of beet canopy nitrogen content.2. To pre-process the beet canopy digital images, image filtering using median filtering, using the H channel threshold-specific method to remove soil disturbance background, using Otsu method eliminates shadow interference background, using Cb channel threshold-specific method eliminates shadow interference background, Finally,we got beet leafy canopy images.Using MATLAB software extract beet canopy R,G,B color data.3. Beet Nitrogen diagnostic analysis are based on digital image information, using the SMLR method established the nitrogen content prediction model, prediction correlation coefficient r=0.84, RMSE=1.63. Set image data principal component analysis,using SVM establish beet nitrogen nutrition diagnosis model, Prediction correlation coefficient r=0.79, RMSE=1.85,and using BP neural network establish model, Prediction correlation coefficient r=0.85, RMSE=1.26. Research indicates, BP neural network model than SVM and SMLR has better prediction.4. Beet Nitrogen diagnostic analysis are based on VNIR information, Convert the spectrum into vegetation index information.Using the SMLR method established the nitrogen content prediction model, prediction correlation coefficient r=0.77, RMSE= 1.09.5. Beet Nitrogen diagnostic analysis are based on VNIR and digital image information fusion,fusion vegetation index information with color characteristic value information, and set data principal component analysis, using SVM establish beet nitrogen nutrition diagnosis model, Prediction correlation coefficient r=0.82, RMSE=1.64, and using BP neural network establish model, Prediction correlation coefficient r=0.88, RMSE=1.03. Research indicates, BP neural network model than SVM has better prediction.