Hyperspectral Remote Sensing Monitoring Of Pinus Densiflora Bursaphelenchus Xylophilus Disease Infection Stages

Bursaphelenchus xylophilus is a kind of invasive pest.With its extremely strong adaptability,it has spread rapidly in China,from south to north,causing huge losses to the country’s ecological economy.Hyperspectral remote sensing has continuous spectral characteristics,which has great significance to the identification of infection stages of Bursaphelenchus xylophilus disease.The research area locates in Dadongtuan Village,Huancui District,Weihai City,Shandong Province.The research object is Bursaphelenchus xylophilus disease and the purpose is to study the hyperspectral remote sensing monitoring methods of Bursaphelenchus xylophilus disease at different susceptible stages.Ground non-imaging hyperspectral data,UAV imaging hyperspectral data and ground measured physiological parameter data were used in this research.The monitoring models were built at the leaf and canopy scales by using Successive Projection Algorithm(SPA)-Fisher discriminant analysis and stepwise discriminant analysis methods based on sensitive bands,spectral indexes and physiological parameters.The main research content and results of the paper are as follows:(1)At the leaf scale,two types of susceptible stage monitoring models were constructed based on the ground hyperspectral full-band(400nm-2400nm)and 400nm-1000 nm bands.One was based on the hyperspectral data characteristic variables,and the other was based on the physiological parameters.By analyzing the sensitivity of different spectral bands to the physiological parameters and different susceptible stages,the appropriate spectral bands for different susceptible stages monitoring were clarified.Based on the ground hyperspectral full-band data,the accuracy of the three susceptible stage monitoring models constructed by the sensitive spectral bands screened by SPA,spectral index and the secondary selection of characteristic variables are 90%,87.5%,and 83.75% respectively;The water content inversion model constructed by sensitive band and spectral index respectively can be used to monitor the mid-term and death stage of Bursaphelenchus xylophilus.Based on the full-band data,the model established by SPA screened bands has the highest accuracy and can be used for monitoring at different susceptible stages.Based on the ground hyperspectral 400nm-1000 nm band data,the accuracy of the three susceptibility stage monitoring models constructed by SPA sensitive bands,spectral index and the secondary selection of characteristic variables were 87.5%,87.5%,and 81.25% respectively;The water content inversion model based on spectral index can be used for monitoring in the middle and late susceptible stages;The results of the model inversion based on the chlorophyll a content established by the spectral index can be used for the monitoring of Bursaphelenchus xylophilus in the middle and later stages,and the recognition accuracy can reach 100%.(2)At the canopy scale,monitoring models for different susceptible stages were constructed by SPA sensitive bands,the spectral index and the second screening of the characteristic variables.The monitoring model of the susceptible stage was constructed based on the corresponding bands to the SPA sensitive band selected by the leaf scale of the UAV hyperspectral data.The results showed that it could only be used to identify dead wood of Bursaphelenchus xylophilus,with an accuracy of more than 80%;The accuracy of the monitoring model based on the UAV hyperspectral SPA-selected sensitive bands was 85.71%,which can be used for monitoring at different stages.Based on the spectral index extracted from the leaf scale,the hyperspectral corresponding band of the UAV hyperspectral was used to build a monitoring model for the different susceptible stages.The model had a very low accuracy and cannot be used for actual detection.Based on the spectral index extracted from the UAV hyperspectral data,a susceptible stage monitoring model was constructed.The accuracy was 88.57%.The model can be used to monitor the three stages of health/non-performance,late susceptibility,and death.In addition,the monitoring model was constructed based on the second selection of characteristic variables extracted from the leaf scale.The model can only be used for the identification of dead wood with an accuracy of more than 80%.Based on the second selection of characteristic variables extracted by UAV hyperspectral data,a monitoring model was constructed.The accuracy of this model can reach more than 85%,and it can be used to monitor the different susceptible stages at the canopy scale.In general,When using the same modeling method,the modeling results based on the ground hyperspectral full band was better than that of 400nm-1000 nm,and the modeling results of direct selection of feature variables based on UAV hyperspectrum was superior to the models of leaf scale direct push up,which laid the foundation for the monitoring of leaf scale and canopy scale of Bursaphelenchus xylophilus disease.