Remote Sensing Monitoring And Mapping Of Cotton Canopy Nitrogen Drone

Nitrogen,which is one of the three major elements in crops,has a great impact on crop growth and soil fertility.UAV remote sensing technology has the following advantages,such as non-destructive,fast and accurate characteristics.UAV remote sensing technology has become a promising method to monitor crop canopy nitrogen in recent years.In this study,the cotton canopy nitrogen of the First Division Irrigation Experimental Station in Alar,Xinjiang was used as the test object.Nitrogen and hyperspectral data of cotton leaves in four periods were collected on July 10(bud period),July 26(flower period),August 2(flower bell period),and August 26,2019(bell period).For cotton leaf nitrogen and hyperspectral data,more than 60 samples were collected in each period.Successive Projections Algorithm(SPA),Standard Normal Variate transformation(SNV),normalization,and data centering methods were used for pretreating the original spectrum.Matlab R2016a software was employed to preprocess the spectral images,filter the sensitive bands.ENVI software was used to extract cotton regions based on cotton canopy multispectral data,and calculate the vegetation index.The cotton nitrogen classification map was made by using the sensitive band of the model,corresponding to the spectrum of the drone.The main research results are as follows:(1)The original spectrum collected on the ground has been preprocessed and the sensitive band has been screened.The correlation analysis method was used to screen four vegetation indexes CIred edge,NDVI,PPR,and GM,which had a high correlation with the nitrogen content of cotton.They were used to predict the nitrogen content in the bud stage,flowering stage,flower bell stage and bell stage,respectively.(2)Partial least squares regression method was used to establish cotton nitrogen prediction models in different periods.The R~2 value of the prediction model is 0.46327 in the bud period,0.47334 in the flowering period,0.66231 in the flowering and belling period,and 0.41725 in the belling period.The RMSE is between 0.1414 and 0.2604.(3)The multi-spectral data of the drone in the cotton fields of the four growing periods were pre-processed by radiation calibration,atmospheric correction,geometric correction,and image cropping.After supervised classification using the Support Vector Machine(SVM)method,the cotton planting area was extracted.(4)The four selected vegetation indices,combined with the nitrogen prediction model,are used to make a map of the nitrogen distribution in the cotton canopy.