Monitoring The Damage Of Brown Planthoppers Based On Physiological Indices And Canopy Spectra Of Rice

Brown planthopper(BPH)is an important pest of rice.Timely and accurate monitoring and prediction is the premise of effective control of brown planthopper.At present,the monitoring of BPH lacks automatic methods and relies on manual investigation.In this study,physiological indices and canopy spectral reflectance of rice exposed to different numbers of BPH were measured,the weight per panicle and total panicle weight of rice in the experimental plot were measured after the rice ripened,and the rice yield was obtained,the damage monitoring model of BPH was established based on the physiological indices and spectral reflectance of rice,so as to provide a method for automatic monitoring of BPH.The main results are as follows:The SPAD value,chlorophyll content,relative water content,soluble sugar content and silicon content of rice leaves in different days after BPH damage were measured,respectively.It was found that the physiological indices of rice leaves decreased gradually with the increase of the number of brown planthopper and the prolongation of the damage time.These indices can be used to establish the monitoring model of BPH number and yield.The relationship between the number of BPH and the yield of rice was analyzed,the number of BPH at booting stage,heading stage and flowering stage was closely related to the yield of rice,it is important for monitoring the number of BPH at booting stage,heading stage and flowering stage.Based on the raw and pretreated spectral reflectance,and physiological indices,the regression monitoring models for the number of BPH and rice yield were established,respectively.The regression monitoring model of BPH,weight per panicle and total panicle weight was established by using the raw spectral reflectance and smoothing,wavelet denoising and multiple scattering correction pretreated spectral reflectance of the rice canopy.The accuracy of the model was improved after the multiple scattering correction pretreated,but the prediction accuracy of regression model is not highPartial least squares regression(PLSR),support vector machine(SVM)and BP neural network(BPN)model for monitoring the BPH number and yield were established based on different pretreated spectral reflectance and physiological indices,respectively.PLSR model has better accuracy in monitoring the number of BPH,The prediction accuracy at booting stage and heading stagein 2019 were 71.4%,and the prediction accuracy of model increased by 28.5%after added physiological indices,respectively.The prediction accuracy at booting,heading and flowering stage in 2020 were 64.3%,78.6%,85.7%,and the prediction accuracy of model were increased by 14.3%,28.6% and 42.8% after added physiological indices,respectively.The monitoring model for the weight per panicle of the damaged rice showed that the prediction accuracy of the BPN model in 2019 was better,and the prediction accuracy of the booting,heading and flowering model were 85.7%,the prediction accuracy of the model increased by 14.3% after adding physiological indices,respectively.The prediction accuracy of the SVM model in 2020 was better,the prediction accuracy of booting,heading and flowering model were 78.6%,85.7%,78.6%,the prediction accuracy was increased by 21.5%,28.6% and 7.2% after added physiological indices,respectively.The monitoring method for the total panicle weight was selected based on pretreated spectra and physiological indices,it found that the BPN model had a better prediction accuracy for total panicle weight in two years.In 2019,the prediction accuracy of booting,heading and flowering stage model were 85.7%,the prediction accuracy of booting,heading and flowering models increased by 28.6%,28.6% and 14.3% after adding physiological indices,respectively.The prediction accuracy of booting,heading and flowering model in2020 was 78.6%,85.7% and 78.6%,and the prediction accuracy was increased by 21.5%,28.6% and 28.6%,after added physiological indices,respectively.In conclusion,physiological indices and canopy spectral reflectance of rice can be used to monitor the number of BPH and yield of rice.The prediction accuracy of the model based on physiological indices of rice and spectral reflectance pretreated by multiple scattering correction was improved,PLSR,SVM and BPN model were more accurate in monitoring the BPH number and yield at booting,heading and flowering stages,and can be used to monitor the number of BPH damage.