Target discrimination is a key step of SAR-ATR, which can remove many false ROI chips obtained by prescreening adopting anomaly detection. A new method of automatic target discrimination was proposed in this paper. Firstly, the results of CFAR detection were discriminated using the area of target, and ROI chips are exported. The features were extracted from the ROI chips, and then the ordered discrimination was realized by comparing the features with the thresholds based on the features. The real X band SAR image data was used to test the discrimination algorithm above mentioned, and finally the discriminated ROIs were presented.