The Research And Application Of Monitoring Method Of Cotton Aphid Hazard Based On Hyperspectral Imaging

Xinjiang is the main cotton producing area in China.In the process of cotton growth,cotton aphid is often damaged,which affects the growth and development of plants,resulting in reduced yield.Traditional monitoring of cotton aphid information requires manual observation of the changes of cotton leaf damage caused by cotton aphid and statistics of the number of cotton aphid,which is timeconsuming and laborious.Moreover,due to the delay of statistics,information of cotton aphid situation lags behind,which is not conducive to the precise prevention and control of cotton aphid damage.The popularization of field sensors makes it possible to acquire aphid information quickly.Therefore,a method of fast identification of aphid harm is explored to improve the efficiency of insect information acquisition.This study based on the current cotton aphid damage monitoring rely on artificial methods of investigation,arrangement of equipment acquisition field caused by the plant a series of problem such as poor access to information timeliness,the coverage of small,health with the hyperspectral imager,victims of cotton aphid leaf samples hyperspectral data and image data,the spectral feature extracting,image characteristics,with a characteristic data as input to construct machine learning model,complete set based on highlights like cotton aphid discriminant model and the harm of cotton aphid damage identification model,and on the basis of cotton aphid damage identification model,based on the android system,independent research and development of cotton aphid damage diagnosis system,To diagnose the harm of cotton aphid and provide technical support for prevention and control of cotton aphid in agricultural production.The main findings are as follows:1.Establishment of damage identification model of cotton aphid based on hyperspectral imagingThis study is based on the new land early after 45 cotton leaf by cotton aphid damage spectral reflectance,the image texture characteristics change,through the different dimension reduction method to extract the spectral characteristic value and texture characteristics,to distinguish machine learning model based on input,aphis gossypii harm discriminant model is set up,implement aphis gossypii hazard leaf recognition based on graph fusion technology.The results show that the rf-pls-LDA model is the best,and the recognition rate of prediction set is 91.49 %.Due to the large amount of model input data,the modeling efficiency is low,and the model is optimized by reducing the model input data.The second-order statistics of texture features were used as input to establish the discriminant model one by one.The PCA-Loading-PLS-LDA-Energy model had the best performance and the recognition rate of prediction set was 92.55 %.2.Establishment of damage degrees of diagnosis model of cotton aphid based on hyperspectral imagingThis study based on hyperspectral technology by extracting health,victims of aphids LuMian research 24 high cotton leaf spectral characteristics,based on hyperspectral feature,the number of aphids damage time,divided into health,mild aphid,severe aphid degree of harm to three aphid,ultimately through the Pearson correlation analysis it is concluded that high spectral characteristics and the three show significant correlation.Study choose the highest correlation aphid damage degree as input indicators,with high spectral index model and partial least square method for the model base,build the cotton aphid damage identification model,a continuous projection of the partial least squares(SPA-PLSR)model to estimate the best accuracy,model decision coefficient R squared is 0.5258,the minimum root mean square error is 0.4021.3.Establishment and application of damage degrees of monitoring system for cotton aphid based on hyperspectral imagingBased on android platform,this study developed a hazard degree identification system based on the hazard degree diagnosis model of cotton aphid.The system takes spectral reflectance of characteristic wavelength as input and harm degree of cotton aphid as output.The system has achieved preliminary results in the practice of shihezi general farm,and realized the rapid monitoring of cotton aphid based on the high spectral imaging technology,which provides technical support for the monitoring of insect situation in farmland.