Research On Extraction Methods For Hyperspectral Characteristics Of Wheat Scab

Wheat scab is an important disease of wheat,and the wheat producing areas with humid and rainy climate are seriously affected.If the disease is not found in a timely,it is likely to cause serious losses to a large-scale epidemic has occurred,resulting in reduced wheat production and even no harvest.,Therefore,it becomes very important for the early prevention and control of wheat scab.At present,the diagnosis of wheat scab is usually carried out by artificial recognition method,and spectral analysis technology has the characteristics of rapid,nondestructive and high efficiency,which is a technical means that can be used to develop crop diseases in a wide range and quickly and effectively.However,the early disease of wheat scab was not obvious,and the spectral characteristics were complex.So far,it is not possible to diagnose and identify wheat scab in early.The overall aim of this current study is to improve the prediction accuracy of wheat scab early identification model.The experiment took normal wheat and wheat at different stages of disease as the research object,which used indoor visible near-infrared(VNIR)imaging hyperspectral analysis of 400-1000nm to analyze the hyperspectral time series characteristics of wheat scab,by selecting different spectral feature extraction methods.The research that focusing on the extraction of spectral characteristics of wheat scab in different periods and the analysis of the timing characteristics of wheat scab and the disease characteristics determination,would make the foundation for the diagnosis and identification of early period of wheat scab.The major contents and results are:(1)A time series feature extraction method of wheat scab based on machine learning was proposed.In order to reduce the influence of independent variables and improve the classification accuracy of the model,the spectrum feature of wheat scab in different periods were selected by SPA,x-LW with RF.And through the spectral characteristic classification model of wheat scab was established by extracting the characteristic wavelength.The results show that the SVM classification model based on the characteristic wavelength extracted by RF has a higher classification accuracy in any period,and is significantly better than the full-band,SPA and x-LW.The validity of RF-SVM model was verified,which laid a foundation for the analysis of wheat disease time series.(2)A time series feature extraction method of wheat scab based on vegetation index was stdied.the spectrum feature of wheat scab in different periods were selected by NDVI,EVI with WDRVI.And through the classification accuracy and Kappa coefficient were analyzed to evaluate the wheat disease characteristics constructed by vegetation index.The results show that the spectral features are best classified by WDRVI.The classification accuracy of later wheat scab was as high as 91.9%and the Kappa coefficient was 0.901,which was better than the spectral feature classification recognition rate constructed by NDVI and EVI.The reliability of the establishment of vegetation index was verified,which laid a foundation for the analysis of wheat disease time series.(3)The characteristics of wheat scab time series distribution and disease determination were given.The hyperspectral characteristics of early period,medium term and late period were analyzed during wheat disease.The distribution and characteristics of wheat spectral characteristics in each period were discussed,and the wheat time series disease was judged.The results show that when the hyperspectral features are mainly concentrated in the red border region,wheat scab is at the early.And the spectral features are mostly concentrated in the range of the "red edge" right boundary of the spectrum and close to the "red edge" right boundary,wheat scab is at the middle.And most of the spectral features are concentrated in the high platform reflection region and exceed the range of the"red edge"right boundary,it indicates that wheat scab at this time is in the late.The research results of this thesis have important theoretical and practical significance for constructing the early diagnosis and identification model of scab,improving the prediction,prevention and control ability of scab and ensuring high yield and quality of wheat.