Identification Of Grassland Species And Prediction Of Chlorophyll Content Based On Hyperspectral Technology

The hyperspectral images of the grassland grasses can be collected by using the hyperspectral spectrometer,which can obtain the image information and spectral information at the same time,which becomes a new method for dynamic monitoring of grassland.In this study,the hyperspectral images of grassland grasses in the range of 400nm?1000nm visible-near infrared spectroscopy were obtained by using the hyperspectral spectrometer.Firstly,the spectral data of the region of interest(ROI)were extracted with ENVI.Secondly,the extracted data were dimensioned and classified again.Finally,the chlorophyll content of the pasture was analyzed.The main contents of the thesis are as follows:(1)The average spectral data of the extracted pasture were pretreated by using Multiplicative Scatter Correction(MSC),Normalize and Normalized Variations(SNP)to eliminate the influence of scattering and increase the spectral absorption information which related to the content of ingredient.(2)30 feature bands were selected to eliminate data redundancy by using Successive Projections Algorithm(SPA),Local Linear Embedding(LLE),Principal Component Analysis(PCA)algorithm.(3)The selected feature bands were classified by Support Vector Machine(SVM),1-Nearst Neighbors(INN)and BP neural network.The hyperspectral image data of 10 classes of 500 samples were used.35 samples among them were used as training set and 15 samples were classified as prediction sets.(4)The BP neural network,partial least squares method and SPA model were used to predict the chlorophyll content of forage.The hyperspectral image data of 3 classes of 150 samples were used in the experiment,35 samples of every class as the training set and 15 samples as the prediction set were used to predict the chlorophyll content of the pasture.The experimental results showed:It was feasible to adopt the hyperspectral imaging technique to identify the grass species of grassland,and SPA was used to extract the spectral characteristic band in the case of 30 characteristic bands could achieve 100%effect when the full range of data was reduced dimension.So the dimension reduction effect of SPA algorithm which could be flexible and easy to select the number of bands was better.Compared with the classification time and effect.SVM achieved good results in judging the grass species of grassland.In the forecast of chlorophyll content,the partial least squares regression model could obtain more accurate prediction results.