A FINE CLASSIFICATION ALGORITHM FOR VEGETATION BASED ON NDVI TIME SERIES FEATURES

The fine classification of vegetation is a prerequisite for relevant remote sensing applications. Based on the phenological differences of different vegetation and the temporal changes of vegetation index, a crop fine classification algorithm is proposed, which combines spectral angle mapper algorithm and spectral information divergence algorithm. Taking Shihezi City, Xinjiang Province as the research area, 15 Sentinel-2 high-resolution images are selected as the data source, combined with the local phenological information, the NDVI time series curves of different crops are constructed, the refined classification of crops in this area is carried out, and compared with the traditional SVM algorithm and maximum likelihood method, The results show that the overall classification accuracy of this algorithm is 89.25% and 86.45% respectively, the kappa coefficient reaches 0.79 and 0.78. The experimental results fully show the advantages of the proposed algorithm compared with the traditional single temporal data source algorithm, and have a certain reference significance for the application of remote sensing images in ground feature classification.