| The vegetation community of The three gorges reservoirs area is the fundamental part of the ecosystem.Besides,the vegetation community composition and the biodiversity in the reservoir area are directly affected by the rise and fall of water level caused by Three Gorge Project.And most of the previous studies on vegetation communities in reservoir areas are based on field investigation more than hyperspectral technology.So this study will break through the traditional method and monitor the vegetation community based on hyperspectral remote sensing,and methods of vegetation community classification were studied to provide important reference for the study of vegetation community in reservoir area.This paper takes the pengxi river basin in the three gorges reservoir area as the research area.Based on the PIKA –L heperspectral remote sensing obtained by unmanned aerial vehicle and relevant field measured data and survey data,and the spectral characteristics of Xanthium sibiricum,Bermuda grass,rice paddy,corn and peanut,a typical vegetation community,are obtained respectively from hyperspectral remote sensing images.In addition,spectral differentiation method is used to analyze these spectral features,and descriptive statistics method is used to select the bands with the largest spectral feature sepration.Furthermore,21 vegetation indexes of each vegetation community were analyzed based on the method of descriptive statistical frequency to select the vegetation indexes with high degree of separation.Finally,a new vegetation index was constructed by combining the spectral characteristic analysis results with the index analysis results,which was applied to the decision tree classification of vegetation communities in typical level-fall zones in the three gorges reservoir area,and the following conclusions were obtained:(1)The vegetation community in the near infrared wave band is higher than the measured spectral characteristics of the average average executive remote sensing image spectral characteristics,but the difference between the measured spectral characteristics of vegetation community class and class of hyperspectral remote sensing image spectral characteristics differences between relatively consistent,based on high spectral characteristics of vegetation community vegetation discriminant research have certain scientific nature.Therefore,it is scientific to discriminate vegetation community based on hyperspectral characteristics of vegetation community.(2)Vegetation spectral characteristic difference is mainly manifested in the green mountain red edge position location near infrared wave band,the band of 549 nm,700.6 nm and 713.5 nm,722 nm,799.9 nm are conducive to the classification of vegetation communities are extracted.(3)The proper selection of the bands used in the construction of vegetation index has an important impact on the index difference of vegetation communities.Simple difference vegetation index and ratio vegetation index have a poor effect on the separation of vegetation communities,while complex ratio vegetation index has a better effect on the separation of vegetation communities.(4)According to the intuitive analysis of the descriptive statistics of each vegetation index,about 81.5% of Siberian Cocklour Fruit on the CI1 index and about 85.6% of corn on the CI2 index are separated from other categories,and about 93.6% of dog roots on the CI3 index are separated from other categories.The CI1,CI2 and CI3 indexes all increase the difference of vegetation communities to some extent.(5)The Kappa coefficient of the newly constructed vegetation index for decision tree classification is 0.60,and the Kappa coefficient of the band classification is 0.55.In general,The accuracy of classification by index is higher than that by band.Therefore,it can also be shown that the vegetation classification using the index constructed in this study has certain robustness(6)The correlation between the spectral characteristics of vegetation communities and vegetation coverage in most bands is higher than 0.7.Moreover,the degree of vegetation coverage affects the classification accuracy,and it is easy to be misclassified into other categories in areas with low vegetation coverage.The newly constructed vegetation index has a better classification effect in areas with high vegetation coverage... |
PDF Full Text Request