Research On Identification Of The Main Plants In The Seriphidium Transiliense Desert Grassland Based On Hyperspectral Imaging Technology

The species composition of grassland community and its variation of growth and decline is an important index to judge the quality characteristics of grassland number and community succession,using remote sensing to identify grassland species will help to improve the quality and effectiveness of grassland monitoring.Based on the main plants of Seriphidium transiliense desert grassland distributed in Xinjiang in China,used SOC710VP hyperspectral imaging spectrometer in April,June and September,three periods of field ground object spectral data acquisition,with Seriphidium transiliense、Ceratocarpus arenarius、Petrosimonia sibirica and bare land as identification objects,based on the analysis of the spectral characteristics and diferences of the three plants,bare land and communities,the band,characteristic parameters and vegetation index three kinds of identification parameters of the four identification objects were screened,and the identification model was established by Fisher discriminant analysis method.In order to provide the method basis for the identification of grassland remote sensing species.The main findings are as follows:(1)The three main plants and communities of Seriphidium transiliense desert grassland showed the spectral characteristics of two peaks and three valleys typical of vegetation,but the reflectivity and reflection amplitude in the bands of 400～1000 nm are different in different periods,and the bare land spectrum tends to be straight,which is obviously different from the vegetation.(2)Among the selected identification parameters,the best characteristic bands were 403.95,418.85,598.68,687.27,698.17,767.13,820.48 and 970.64 nm in June;The best characteristic parameters were Lb、Lr,Are,Lb、Mb、Mg、Lr、Mr、Ab and Are,Lb、Lr and Are in September,respectively.The vegetation index in June was the best.The identification parameters of Seriphidium transiliense,Ceratocarpus arenarius and bare land were all NDVI、RVI、DVI、BND VI、GND VI、EVI、NDGI、RI、PSRI and VOG,while were ND VI、RVI、DVI、BNDVI、GNDVI、EVI、NDGI、RJ、PSRI、RENDVI and VOG.(3)Using the characteristic band,characteristic parameters and vegetation index discriminant model established by Fisher discriminant analysis,the total accuracy of vegetation index(90.37%)>feature parameter(85.98%)>feature band model(83.15%),vegetation index(97.88%)>characteristic band(92.27%)>characteristic band(88.48%)in June,vegetation index(91.53%)>characteristic band(89.77%)and characteristic band(84.87%)in September.There are certain differences in the spectral characteristics between the identification objects,and the selectivity of the identification parameters is different in different growth periods;the overall accuracy of the identification model reaches more than 80%,and the model established by the vegetation index is the best identification model.