Forests play an important role in the global carbon cycle and natural air conditioning. Monitoring and mapping of forest distribution are of great significance. With the successive launch of new synthetic aperture radar (SAR) sensors, microwave remote sensing data acquisition methods have been developed from single-band, single-polarization and single-angle to multi-frequency, multi-polarization, multi-angle, multi-temporal and so on. That provides an unprecedented potential and opportunity for SAR in the research and application of forest identification. In this paper, the data source mainly included the quad-polarization C-band GaoFen-3(GF-3) and dual-polarization L-band ALOS-1 PALSAR. First, the single-look complex (SLC) data was preprocessed with multi-look, filtering, radiation calibration, geocoding, registration and clipping. Three polarization characteristic parameters of entropy (H), scattering angle (α) and anisotropy (A) were obtained by using Cloude-Pottier polarization decomposition, and three texture features of the mean (MEAN), variance (VAR) and dissimilarity (DIS) were extracted based on the gray-level co-occurrence matrix(GLCM). Combined with the advantages of GF-3 high-resolution quad-polarization and PALSAR L-band, multi-dimensional information including frequency, polarization, temporal and texture features was used synthetically. Then support vector machine (SVM) supervised classifier was used to obtain the four classification results, including coniferous forest, broad-leaved forest, mixed broadleaf-conifer forest and others. The experimental result shows that proposed method achieved a better classification result based on multi-dimensional POLSAR, the overall accuracy of forest type identification is approximately 89.47% and the Kappa coefficient is 0.85.