Polarimetric Synthetic Aperture Radar (PolSAR) data classification methods, which are proposed in various studies, useing the information of one polarization base only. In this paper a new method is presented for classification of polarimetric SAR data is presented that this method can use more than one polarization base information. The proposed method is based on polarimetric signatures. In this study, in addition to the polarimetric power signatures, polarimetric signatures were introduced for various PolSAR features were introduced. As the other knowledge-based classification methods, the proposed presented method has two steps. First, extracting reference knowledge and second, classification of data using reference knowledge. Polarimetric signatures for various PolSAR features were used for extracting reference knowledge. Also, pattern recognition matching algorithms were used for classification. To the produce of polarimetric signatures in the segments and to avoid the noisy output, the object-based method was used. A Radarsat-2 image of Petawawa forest area was chosen for this study. According to the results, in comparison to the accuracy of wishart classifier which is 76.34 percent, the accuracy of proposed offered method is 82.12. The aAccuracy of forest species had a significant improvement in the proposed suggested method. The method was successful because of three major factors: fFirst, using a more complete set of polarimetric features., Ssecondly, increasing the feature information using polarimetric signatures, and thirdly, employing contextual information in an object-oriented method.