Mangrove species classification based on multiple vegetation index extraction and joint sparse representation

Mangrove species classification is important for studying the changes of mangrove ecosystem.In this paper,the distribution of mangroves in Tieshangang is chosen as the study area,and domestic ZY3 mapping satellite data is adopted as the data source.The spectral characterization of various mangroves is analyzed,four different vegetation indices (RVI,NDVI,VARI and NDGI) are exacted to add vegetation information,and the joint sparse representation algorithm is utilized to distinguish seven mangrove species.We analyze the vegetation index of seven mangroves (Aegiceras corniculatum,Excoecaria agallocha,Avicennia marina,Rhizophora stylosa,Kandelia candel,Sonneratia apetala and Bruguiear gymnorrhiza) and other species (bushwood,mudbank and grassland).The geometric dimension and spectral dimension are combined and the joint sparse representation is used for classification.The overall accuracy reached 95.37% and kappa coefficient reached 0.9347 when we use the spectral data incorporate with four vegetation indices.Experiments show that spectral features combined with vegetation indices can improve classification accuracy,and NDVI has a greater contribution than other indices to distinguish mangrove species.Furthermore,joint sparse representation classifier has a good performance in mangrove species classification.