Hyperspectral Remote Sensing Monitoring Of Scab Of Winter Wheat Based On Different Scales

In recent years,with the global warming,the prevalence of scab is increasing year by year.Wheat infected by scab will be poisoned after human and livestock eat.Therefore,the prevalence of scab will seriously restrict the yield and quality of wheat in China,and increase the economic loss of farmers.How to diagnose scab quickly and accurately will become very important.Remote sensing monitoring technology is characterized by real-time,objective and accurate,especially hyperspectral remote sensing,which can provide important technical support for disease assessment in crop pest monitoring research.In this thesis,two different scales of wheat and canopy were studied.The spectral data of wheat ears and canopy were collected by ASD FieldSpec Pro,combined with machine learning related methods,the disease regression and disease degree of scab were studied to provide technical support for disease evaluation.The main research contents and results of this paper are as follows:(1)Hyperspectral remote sensing monitoring of scab of winter wheat based on ear scaleTaking the wheat ears of infected wheat as the research object,using ASD hyperspectral data,on the one hand,by analyzing the severity of spectral differentiation and scab,the two sensitive bands of 450-488nm and 500-540nm were selected,and then The normalized ratio of the first-order differential sums in the two sensitive bands ranges to construct the Wheat Scab Index,On the other hand,a linear regression model was constructed by using 11 vegetation indices and 12 spectral differential indices(including WSI index)and severity of disease.The results show that the coefficient of determination(R2)of the WSI model is 0.73,which is better than the results of other vegetation indices(The NBNDVI model has the highest coefficient of determination,R2?0.67)and other differential indices(the SDy model has the highest coefficient of determination,R2=0.64).Finally,the 11 vegetation indices and 12 spectral differential indices were used as independent variables,and the severity of the disease was multivariate stepwise regression.The effect(R2=0.79)was better than the one-dimensional linear regression constructed by single index,which can provide methods and model support for nondestructive monitoring of wheat scab.(2)Hyperspectral remote sensing monitoring of scab of winter wheat based on canopy scaleTaking the hyperspectral data of wheat canopy as the research object,the spectral differential features and wavelet features are extracted by first-order differential and continuous wavelet transform,as well as the vegetation index,which is often used for disease monitoring.Among the three types of features,the features that are significantly correlated with the severity of scab are selected as sensitive features of scab.Among them,feature extraction based on wavelet analysis has a great advantage in disease information extraction.Based on the original spectral data and sensitive features,the disease monitoring models of support vector machine(SVM)and random forest(RF)were constructed,and the disease monitoring model of PCA-Bayesian was constructed based on the original spectral data.The results show that:in the SVM disease classification model,the model whose input variable is spectral feature has better classification effect(76%)than the original spectral data(63%),indicating that the sensitive feature improves the classification accuracy of the SVM model.In the disease classification test based on PCA-Bayesian,the classification effect of PCA-Bayesian(83%)is better than that of SVM model.In the RP classification model,the RP classification effect(89%)is better than that of SVM(76%)and PCA-Bayesian(83%).