Abstract: The remote sensing image digital of SPOT-5 of Saihanba Mechanized Forest Farm was being adopted distinguishing object in this paper. The three different classifiers of supervised classification— maximum likelihood classifier, minimum distance method and parallelepiped method were being adopted. The results showed that classification accuracy of maximum likelihood classifier was 86.25% and it was the best suitable method for the classification of arbors in Saihanba area. The classification accuracy of maximum likelihood classifier was 14.73 percentage points more than minimum distance method and the parallelepiped rule was not according with practical condition in this research.