Urban vegetation classification based on phenology using HJ-1A/B time series imagery

Abstract

Urban vegetation classification need vegetation index especially temporal information of vegetation, thus high spatio-temporal NDVI product is necessary. NDVI time-series data derived from HJ 1A/B time series imagery (HJ NDVI) have relatively high spatio-temporal resolution. In this research, HJ NDVI time series of typical vegetation types in the city of Nanjing are established, the S-G filter is chosen to filtering. Taking filtered HJ NDVI time-series data as “simulated Hyperspectral data”, the linear spectral mixture unmixing algorithm is used to carry out vegetation mapping. The results indicate that unmixing algorithm of linear spectral mixture model can obtain the distribution information of the five kinds of vegetation sub-classes including shrub, grassland, evergreen needle forest, broadleaved deciduous forest, evergreen and deciduous broad-leaved mixed forest in the research area.