Monitoring And Classification Of Wheat Take-all In Field Based On UAV Hyperspectral Image

Wheat is one of the most important food crops in China.In recent years,wheat pests and diseases have occurred on a large scale in some areas of China,resulting in reduced winter wheat production and reduced quality.With the emergence of remote sensing technology and the need for precise management of wheat crops at regional scale,the use of remote sensing technology to obtain,process and monitor the occurrence and development of wheat pests and diseases has become an effective means,especially in recent years,the development of hyperspectral remote sensing monitoring technology.Compared with traditional spectroscopic spectrometers,each pixel in a hyperspectral image obtained using this technique contains unique spectral data,which achieves the effect of “map and spectrum integration” of remote sensing data and can not only cause wheat diseases and insect pests.The water content,nitrogen content,chlorophyll content,and tissue morphology of the leaves produce spectral responses,and can also provide a better spatial visualization of the spectral characteristics exhibited by the wheat leaves,providing an important basis for quantitative remote sensing monitoring of wheat pests and diseases.Research basis.In this study,the Wheat Take-all was taken as the research object,and the level of disease monitoring of Wheat Take-all was the main research content.Large-scale,realtime and non-destructive monitoring of the extent of Wheat Take-all is important for its prevention and control.Canopy spectral reflectance obtained by the ASD nonimaging spectrometer on the canopy scale of the winter wheat in the study area and the UHD185 imaging spectrometer mounted on the UAV platform is used.Combined with the ground survey data,a monitoring model for the level of Wheat Take-all was established in this area to provide timely and accurate guidance for the prevention and control of Wheat Take-all.This paper uses ASD spectral data to evaluate the accuracy and reliability of UHD185 spectral data,and then compares the corresponding wheat canopy spectral response differences under different disease index of Wheat Take-all,and builds a hyperspectral remote sensing monitoring model of Wheat Take-all.And use independent sample data to test the reliability and accuracy of the monitoring model.The results showed that the ASD spectral data of winter wheat canopy were significantly correlated with UHD185 spectral data,and the determination coefficient R2 was greater than 0.97.The UHD185 imaging hyperspectral data was selected to be the most reliable(462-874 nm)range(4~102 bands).First,a comprehensive analysis of the difference spectral index(DSI),ratio spectral index(RSI),normalized difference spectral index(NDSI),and the quantification of disease index of Wheat Take-all in the range of 462 to 874 nm Relationships,to construct a monitoring model of disease severity of Wheat Take-all,the analysis shows that the linear regression model constructed by difference spectral index DSI(R818,R534)and wheat disease index of Wheat Take-all has a high correlation(decision coefficient R2=0.8605,Root mean square error(RMSE=0.073,number of modeled samples n=20).Based on the Wheat Take-all monitoring model,independent model data were added to verify the model.Tests showed that the predicted value of the disease index of Wheat Take-all was The measured values have a high correlation(R2=0.76,RMSE=0.149,number of test samples n=20).Secondly,classification methods based on radial basis function,polynomial,Sigmoid and linear kernel support vector machines(SVM)are used to classify hyperspectral images in the study area.The classification results are verified by real reference sources,and radial basis function,polynomial,Sigmoid,and linearity are compared.The classification accuracy of the kernel functions was evaluated.The support vector machine method based on radial basis function was found to have the best classification effect on Wheat Take-all.The classification accuracy reached 90.35% and the Kappa coefficient was 0.86.Compared with the spectral index The method has a higher classification accuracy for the Wheat Take-all.Therefore,based on the radial basis function support vector machine method,we can effectively monitor the disease level of Wheat Take-all.This paper provides the research basis for the application of UAV hyperspectral remote sensing technology in the precise monitoring and application of Wheat Take-all.