Identification And Classification Of Typical Wetland Vegetation In Poyang Lake Based On Spectral Feature

As an important part of the wetland ecosystem,wetland vegetation plays an important role in maintaining and carrying critical physical processes and biological functions of the system.The application of remote sensing technology to the effective identification of wetland vegetation is conducive to the realization of large-scale and high-age monitoring,and provides scientific basis for human protection of wetland vegetation resources and restoration of ecological environment.This study takes the typical wetland vegetation in the Nanjishan Wetland Nature Reserve of Poyang Lake as the research object.Based on the measured hyperspectral data and multisource remote sensing image data,the spectral characteristics of the wetland vegetation are analyzed and identified.Based on field surveys of typical wetland vegetation in different seasons,the basic phenological characteristics of typical wetland vegetation such as Cynodon dactylon,Phragmites australis,Polygonum hydropiper,Carex cinerascens,and Triarrhena lutarioriparia were determined,and the spectral reflectivity information of the canopy was measured in the spring and autumn periods.First,spectral analysis method was used to analyze the spectral characteristics information of typical wetland vegetations in different seasons.Then,spectral bands based on the error range were selected to select spectral bands in different growing seasons and to identify the best recognition season.At the same time,the spectral characteristics of different types of satellite sensors in typical wetland vegetation are analyzed and compared,and based on the band response function,the five planting indices of Landsat OLI,GF-1 WFV,and HJ-1A HSI are simulated(DVI,RVI,NDVI,HJVI,SAVI)Wetland vegetation recognition performance,to identify satellite imagery and vegetation index with better recognition effect;Finally,using appropriate remote sensing images,based on prior knowledge and decision tree classification methods for protected wetland Vegetation clusters were classified by remote sensing.The results show that:(1)The canopy spectral reflectance of typical wetland vegetation in the Nanjishan Wetland Nature Reserve shows obvious differences in different seasons,and the wetland vegetation can be effectively identified on the canopy scale.(2)According to the measured spectral characteristics of the selected hyperspectral bands,it mainly focuses on the red-edge and near-infrared bands,and accounts for the lower proportion of the entire waveband,achieving an ideal spectral dimension reduction effect.(3)The discriminant analysis of wetland vegetation in different seasons based on the selected spectral characteristic bands reveals that the recognition effect in autumn is obviously better than that in spring.(4)Simulation of different satellite sensor experiments found that recognition of wetland vegetation RVI,NDVI and SAVI based on vegetation index showed strong recognition ability,and the recognition effect was better in GF-1 WFV image.(5)Adopting the prior knowledge and combining the CART model's decision tree method to perform remote sensing classification experiments on the dominant wetland vegetation clusters in the protected area,and obtaining higher classification accuracy.