A new classification method based on the support vector regression of NDVI time series for agricultural crop mapping

Time series of remotely sensed data are usually an important source of information for various agricultural applications. However, modeling and analyzing of time series data is not straightforward. In this paper, a new method for classification of time series data is proposed. The method is a modified Maximum Likelihood (ML) algorithms that uses Support Vector Regression (SVR) algorithms to estimates the probability of each class. The method is tested on the NDVI time series obtained from TM and ETM+ data for an agricultural region in Qazvin province, Iran. Six different crops have been considered for this study, and the classification results are evaluated using the kappa coefficient and the overall accuracy. The evaluations of experimental results and their comparison with the classic ML algorithm show that the proposed method, with kappa coefficient greater than 0.9, is a promising method and could be an alternative approach for agricultural classification needs.