Land cover classification with Support Vector Machine applied to MODIS imagery.

The reported study is the first part of an on-going work which objective is to produce a land cover classification of continental Portugal from multi-spectral and multi-temporal MODIS satellite images acquired at a 500 m nominal resolution. Our goal is to achieve an automatic pixel level classification using a Support Vector Machine (SVM) learning approach. More precisely, we use the time evolution of reflectances measured in several spectral bands from weekly composited images acquired during a complete year period. As temporal profiles are relevant fingerprints of local phenologies, we believe time series offer great potential to improve discrimination among the different land cover types. In order to reduce the input space dimensionality, we endeavor at identifying a parsimonious set of fitting parameters that adequately model the time series. Eventually, our model parameters will be used as inputs of a supervised SVM classifier. Performances will be exhaustively compared to those obtained when the same classifier is directly applied to a single date multi-spectral reflectance data. It is this preparatory study demonstrating the importance of the chosen analyzing date that is at the core of the present article. In particular, we show that global misclassification error reaches a minimum at a specific period of the year. However, when each class is taken separately, this date does not necessarily correspond to the most favorable time for its identification.