Application of Hybrid Classification Method Based on Fourier Transform to Time-Series NDVI Images

Fourier transform is presented and applied to monthly composited NDVI data over one year derived from MODIS to examine the frequency distribution of the multi-temporal signal. It is shown that frequencies of the time series of NDVI are linked to integrated NDVI and seasonal variabilities of different periods of the land cover types. The 1/12 month-1 frequency component summarizes the relative dominance of annual habit of land cover types over Hubei province. The introduction of these characteristic phenology parameters extracted from time series of NDVI into feature space improves the separabilities. This research indicates that Fourier transform provides a useful tool for handling temporal sequences of MODIS data and studying the vegetation phenology.