| Hyperspectral and Li DAR technology have become important means for monitoring diseases and insect pests in agriculture,forestry and grassland.The hyperspectral data contains abundant spectral information of pests and diseases,but the spatial three-dimensional structure information is insufficient.Li DAR data has high spatial resolution and rich three-dimensional structure information,but the characteristic information of diseases and insect pests is not obvious.To combine Li DAR and hyperspectral data to extract information of forest diseases and insect pests quickly and accurately,which realize the monitoring and early warning of diseases and insect pests is an urgent problem to be solved at the present stage.Pinus yunnanensis is the main tree species in southwest China,which has wide distribution and large area.Tomicus spp.is a worldwide pest to forestry,which dose great damage to P.yunnanensis.It is difficult to control insect outbreaks and poses a serious threat to ecological security.Therefore,monitoring and early warning of P.yunnanensis damage by Tomicus spp are particularly important.This paper,taking P.yunnanensis damage by Tomicus spp as the research object,carried out correlation analysis between canopy dieback rate and ground survey of whole tree dieback rate;extracted and analyzed hyperspectral features from needles to canopy scales at different damage levels(health,mild,moderate,severe),and obtained damage characteristic curves and the optimal monitoring ban extracted and classified window,and selected the optimal feature extraction and classification method based on ground hyperspectral,airborne hyperspectral and airborne Li DAR technologies.This method combined with airborne Li DAR and hyperspectral data was used to classify and diagnose the damage degree in each P.yunnanensis,and visualize the results.The results are as follows:(1)The remote sensing monitoring data are the characteristic data middle and upper layers of the canopy.In this paper,the correlation between the canopy dieback rate and the whole tree dieback rate of the ground survey was analyzed,showing a highly significant positive correlation.The unitary linear fitting was carried out between the canopy dieback rate and the whole tree dieback rate in the ground survey,and the fitting equation is y=1.011 x,which realized the estimation of the whole tree dieback rate through the canopy dieback rate.And according to this,the damage degree in P.yunnanensis by Tomicus spp was classified.(2)Spectral characteristics were extracted and analyzed from needles to canopies,and the result showed that the damage spectra of Tomicus spp.were similar at the two scales.The spectral reflectance curves of different damage degree were different in "red valley"(640～700nm),and fluctuated significantly in the near infrared band,The spectral reflectance of P.yunnanensis needles decreased significantly as the damage degree by Tomicus spp.increased;The optimal spectral reflectance monitoring band mostly belongs to the red band with the range of 765-838 nm.The first derivative of the spectrum has an obvious crest at the "red edge"(680nm ～ 740nm),which reaches the highest peak.With the aggravation of the damage degree,the peak height gradually decreases,and the "red edge" position is slightly shifted to the direction of blue light.At 750-950 nm,there were obvious peaks and troughs,and the overall trend of the four damage levels was similar.The optimal first-order derivative monitoring band was mainly 702-744 nm in the visible red band.As the damage degree by Tomicus spp increased,the spectral reflectance decreased significantly in the near-infrared band,and increased at the "red valley".When the damage was severe,the "red valley" showed signs of disappearance.The peak value of the curve of the first derivative of the spectrum decreased gradually with the increase of damage degree at the "red edge",and the position of the "red edge" tended to shift to the blue light band slightly.(3)Three feature extraction methods including principal component analysis,vegetation index and continuous wavelet transform were used to extract the characteristics of coronal hyperspectral data of P.yunnanensis collected by UAV with different damage degree.Extract spectral data characteristics after preprocessing,and the extraction results were classified into four classification algorithms: discriminant analysis,support vector machine,nearest neighbor and BP neural network.The classification results were compared and analyzed and the accuracy was evaluated.It was concluded that the CWT based BP neural network classification algorithm had the best diagnostic effect.(4)In order to solve the problem of "foreign matter of the same spectrum,foreign spectrum of the same matter",airborne Li DAR and hyperspectral data were integrated to obtain the canopy pattern spot information of a single plant with automatic single-plant separation.Extraction and combination the hyperspectral data within the pattern spot,can obtain the canopy vector data of a single P.yunnanensis containing hyperspectral information in the forest region.The BP neural network classification algorithm based on CWT was used to classify,diagnose and visualize the damage degree in P.yunnanensis by Tomicus spp.at the regional scale.The overall classification accuracy was 90.83%.Finally,the disaster analysis and judgment of the small class in the experimental area are carried out and the classification results are visualized.In conclusion,in this paper,the damage degree in P.yunnanensis by Tomicus spp.was classified by canopy dieback rate;hyperspectral feature of P.yunnanensis with different damage degree by Tomicus spp.was extracted,classified and diagnosed from needles to canopy scale;data fusion of airborne Li DAR and hyperspectral were used to classify and diagnose the damage degree of a single plant.Some research results from these three aspects have certain theoretical significance and practical application value. |
PDF Full Text Request