A simple cleaning procedure for improvement of training site statistics

For improving the effectiveness of supervised training, a cleaning procedure which operates by selectively dropping training site pixels based on the Mahalanobis distance and class probability has been proposed. The method is iterative and takes into account the spectral overlap in all image hands. The procedure results in greater classification accuracy with a narrower confidence interval. The Bhattacharrya distance measure of class separability improved from an average value of 1.9373 to 1.9797 with a maximum change for a class pair from 1.2671 to 1.9052. The overall classification accuracy increased from 94.74/spl plusmn/0.64 to 39.63/spl plusmn/0.19.<
>