In order to give a comparative estimation of the proposed method, the well-used method of PCA followed by PLS-DA were conducted to have pixel-wise identification accuracies. In detail, training samples were same as in the learning images. All bands were fed to PCA for feature reduction, and first 3 PCs were retained for further PLS-DA classification. The result gave 94.47%, 90.53% and 96.02% for breed A, B and C respectively. Compared with the accuracies, it is obvious that the proposed method in this study obtained better results. Although PLS-DA is good enough, better trained classifier is more desirable so as to be applied in practical detection