Monitoring Nitrogen Nutrition Of Winter Wheat With The Integration Of Unmanned Aerial Vehicles And Satellite Remote Sensing Imagery

Real-time,rapid,non-destructive and accurate assessment of nitrogen nutrient status of crops is important for guiding crop nutrient diagnostic research and precise variable fertilization.Satellite remote sensing,with its advantages of large scale,rapid and informative,has become the mainstream agricultural remote sensing tool at present.However,without sufficient and suitable reference data,the great potential of satellite remote sensing cannot be fully utilized.Satellite remote sensing data often suffer from large-scale scale mismatch and in situ sampling bias when applied to ground sampling data for modeling,and cannot obtain the spatial distribution of nitrogen nutrient status on a large scale.In contrast,unmanned aerial vehicles（UAV）remote sensing meets the needs of high-precision and real-time quantitative inversion for precision agriculture by virtue of its time-sensitive,high spatial resolution and reusability.Therefore,this paper uses the surface information obtained from UAV remote sensing data instead of point sampling information,and applies satellite remote sensing images to estimate the nitrogen accumulation at regional scale.Based on the integrated application of UAV and satellite remote sensing,this paper combines the ground-measured wheat nitrogen nutrient data,and focuses on the relative radiometric correction method of multi-flight UAV images and the integrated construction of the UAV-satellite remote sensing wheat nitrogen nutrient estimation model,respectively,in order to achieve the use of UAV images as an intermediate bridge of simultaneous satellite-ground observation and improve the accuracy of the remote sensing estimation model of wheat nitrogen nutrient parameters at regional scale.To resolve the radiometric differences between multi-site and multi-flight UAV images and to reduce the spectral response differences between UAV and satellite sensors,this study proposes a relative radiometric correction method（MACA）for multi-flight UAV images based on contemporaneous satellite images.The method first applies a machine learning method to establish a multiple subset multiple linear regression model based on the resampled UAV and satellite images reflectance to obtain reference images with satellite reflectance characteristics in the UAV images coverage area;then the least squares regression method is applied to establish a radiometric correction model for each waveband of the UAV image and the reference image to obtain the relative radiometric corrected UAV images.The new method is compared qualitatively and quantitatively with the previously proposed Mean of Ratio,Pseudo-invariant Features,and rule-and example-based regression model method（Cubist）.The results show that the four relative radiometric correction methods reduce the radiometric differences among the multi-flight UAV images to different degrees,and the MACA method has the best correction effect considering the spectral characteristics of the features in the study area,the degree of change of the relative radiometric correction methods on the UAV images,and the ease of operation.Based on this correction method,we first performed relative radiometric correction on multi-flight UAV images and applied vegetation index set,combined with machine learning methods to obtain a spatial distribution map of wheat N accumulation based on the resolution of UAV images.This spatial distribution map is then used to train a machine learning model of vegetation index set at the liter scale,construct a UAV-satellite N accumulation estimation model,and map the spatial distribution of N accumulation at the regional scale.Finally,we evaluated the accuracy of the nitrogen accumulation model based on the satellite-machine remote sensing images.The results showed that the relative radiometric correction could improve the accuracy of N accumulation estimation from multi-site and multi-flight UAV images;The independent validation accuracy of the UAV-satellite plant N accumulation estimation model(R²=0.52,RMSE=26.08 kg·ha^-1)was better than that of the satellite remote sensing estimation model obtained by direct application of ground sampling data(R²=0.07,RMSE=49.33 kg·ha^-1).This paper introduces a research method for nitrogen nutrient monitoring of wheat with the integration of UAV and satellite remote sensing,firstly,a multi-flight UAV image relative radiometric correction method is proposed;then the distribution map of nitrogen accumulation estimated by the UAV model is applied instead of the point information of ground sampling,and a UAV-satellite remote sensing-based model for estimating nitrogen nutrient parameters of wheat is constructed,which improves the accuracy of inversion of nitrogen accumulation by satellite images.The application of the integration of UAV and satellite remote sensing clarifies the performance of the UAV estimation model at different spatial scales,establishes the estimation model at regional scale,and finally achieves the high accuracy estimation of wheat nitrogen nutrient parameters by applying satellite images.In the future,a combination of high spatial resolution and low and medium spatial resolution satellite images will be used to achieve rapid and accurate monitoring of crop growth parameters in the province or even on a larger scale.