For each intensification, the row differences have been computed.
The correlation coefficient matrix for each differenced intensification has been computed, and vectorized as in [1].
The vectorized correlation coefficients for 12 intensifications have been concatenated to form a vector of length 24192.
The training dataset is represented in a 1275 by 24192 matrix, and SVM classification is performed on this matrix.
The feature subset selection has not been utilized.