Research Of Cotton Aphid Monitoring And Platform Based On Image Fusion Technology

Cotton is an important material and strategic resource in China,and Xinjiang is the most important cotton-producing area in China.Cotton is extremely vulnerable to pests during its growth,which seriously affects its yield and quality.This thesis combines image recognition technology,deep learning technology,unmanned aerial vehicle multi-spectral remote sensing technology,atlas fusion technology and other technologies to achieve rapid and accurate monitoring of the occurrence and damage degree of cotton aphids in the cotton field at the field scale,and constructs a visual cloud platform for rapid cotton aphid monitoring,breaks the traditional manual investigation methods,improves the timeliness and accuracy of cotton aphid monitoring and early warning,and reduces the difficulty of cotton aphid sampling survey,The scientific and practical importance of technical assistance in the control of cotton aphids is immense,aiding in the advancement of precision agriculture,guaranteeing the yield,quality,and security of cash crops,and furthering the progress of modern agriculture.The main research results are as follows:1.Construction of cotton aphid damage monitoring model based on UAV RGB imaging.The color index,texture feature and depth feature of UAV RGB imaging data are extracted,the RGB imaging feature data set is established,and multiple discrimination models such as LR,RF,PLS,SVM,CNN,Res Net,Sim AM-CNN,Sim AM-Res Net are constructed,and good monitoring results are achieved.The optimal model was selected for cotton aphid damage inversion,and the accuracy of training set,verification set and test set reached0.7778,0.6875 and 0.6875.2.Construction of cotton aphid damage monitoring model based on UAV multispectral imaging.The full band data,vegetation index and characteristic band of the UAV multi-spectral imaging data are extracted,and the multi-spectral imaging characteristic data set is established,and multiple discrimination models such as LR,RF,PLS,SVM,CNN,Res Net,Sim AM-CNN,Sim AM-Res Net are constructed,and good monitoring results are achieved.The optimal model was selected for cotton aphid damage inversion,and the accuracy of training set,verification set and test set reached 0.7708,0.7083 and 0.6875.3.Construction of cotton aphid damage monitoring model based on image fusion.Using RGB imaging and multispectral imaging data,the feature fusion scheme is adopted for feature fusion.A variety of discriminant models,such as SVM,CNN,Res Net,Sim AM-CNN,Sim AM-Res Net,were constructed using15 fusion feature levels,and achieved good monitoring results.The optimal model was selected for cotton aphid damage inversion,and the accuracy of training set,verification set and test set reached 0.8681,0.8125 and 0.7916.4.Construction of cotton aphid rapid monitoring platform based on UAV imaging.Focusing on the collection,reporting,summary,monitoring,analysis and visualization of the occurrence of cotton aphid pests,a series of applications have been provided for the monitoring,management and control of cotton aphid based on artificial intelligence technology,and the cotton aphid precise monitoring and early warning system of "data collection → data management → data analysis → release of pest results" has been formed,realizing the accurate and rapid monitoring of cotton aphid.