Multi-scale Remote Sensing Monitoring Of Leaf-eating Insect Pests Of Oak Tree Species

Oak is the largest and most widely distributed tree species in Henan Province,with a variety of economic and ecological value,has an important position in the forest ecosystem of Henan Province.In recent years,the oak moth,oak moth,oak moth,oak moth,oak moth,cork oak moth,etc.are the main pests of oak leaf-eating frequent disasters,annual occurrence of more than 60,000 hm2,seriously affecting the economic development of forestry and forest ecological stability.Oak species suffer from insect pests,tree branches on the young leaves,old leaves and leaf veins are eaten resulting in the loss of leaves,in serious cases resulting in oak trees die.Therefore,timely access to the spatial location,disaster area,and pest level of the damage at the early stage of the pest is the urgent need to prevent and control oak pests.Traditional manual survey methods often cost a lot of manpower and material resources,can not achieve rapid monitoring of a large area,easy to miss the best control time.Aerial remote sensing technology to achieve a large area,rapid monitoring of oak insect pest occurrence area to provide important technical support.At the same time,because the UAV remote sensing monitoring technology has the characteristics of rapid,non-destructive,time-efficient and dynamic continuity,can provide technical support for the accurate monitoring of the degree of oak pest damage.Based on "satellite remote sensing-UAV multi-spectral remote sensing-ground survey",this paper uses the multi-scale remote sensing data to remotely monitor and validate the accuracy of the remote sensing model for the identification of the location of oak leafminer pest and the degree of oak leafminer pest on the scale of sample sites.Accuracy verification,the construction of large scale oak pest occurrence location spatial distribution pattern and oak leafminer damage recognition of remote sensing monitoring model.The main research results are as follows:1)Insect pest location identification based on satellite remote sensing at large scale.Based on Landsat-8 satellite remote sensing images,the sensitivity of raw bands and texture features to oak forest pest areas was analyzed,and the results showed that B4,B5,B6 raw bands of Landsat-8 and five texture features of contrast,second-order moments,variance,correlation and mean could reflect better results in delineating affected oak forest areas.After constructing the vegetation index features and using ANOVA to initially screen 22 feature parameters into XGboost classifier for importance analysis,the importance features were sorted and then inputted into the feature parameters in order,and the results showed that the XGboost classifier was used to detect oak infestation areas with 9 feature parameters input,the accuracy of XGboost classifier reached the maximum with OA of 0.915 and The method has certain advantages in the satellite image of oak leafminer pest area delineation,can effectively compensate for the efficiency of manual visual interpretation of the image,in the occurrence of oak pest from the satellite image to achieve a large regional range of initial determination of the pest range,can provide reference for efficient and rapid research on the identification and location of oak leafminer pest location.(2)The classification of pest damage level at the sample site scale based on the multispectral images of UAV.Based on the location of the pest area in the large-scale satellite image,we selected the pest sample sites and obtained the multispectral image data in the pest sample sites based on the remote sensing platform of UAV multispectral data,and classified the pest level by maximum likelihood method(MLC),object-oriented segmentation(e Cognition),support vector machine(SVM)and random forest(RF),respectively.The data were collected with the main spectral data of oak species with different insect damage levels in the sample plots,and the single-band insect-infested tree reflectance was extracted and analyzed for its spectral and textural characteristics and sensitive bands.To reduce the redundancy of classification machine learning,ANOVA was used to filter out the insensitive features to oak leaf-eating insect pests,and finally normalized vegetation index(NDVI),canopy chlorophyll content index(CCCI),greenness normalized vegetation index(GNDVI),normalized difference greenness index(NDGI),mean(Mean)and contrast(Constrast)were selected as optimal features into the machine learning algorithm.The results showed that the overall accuracies of MLC,e Cognition,SVM and RF were 77.0%,82.8%,86.2% and 90.8%,respectively,and the Kappa coefficients were 0.693,0.770,0.816 and 0.877,respectively.RF model had the best classification effect and could be combined with the spectral remote sensing data for the optimal pest classification.At the same time,the spatial distribution map of oak leaffeeding insect pest classes is drawn,which can accurately monitor the damage degree and area of oak pests and provide scientific guidance for oak leaf-feeding insect pest control,resource conservation and management.