A new method based on geometrical features is proposed for aircraft target interpretation to make full use of the characteristics of highresolution SAR (synthetic aperture radar) images. Our method consists of three steps. First, the local selfsimilarity and DBSCAN (densitybased spatial clustering of application with noise) algorithm are used to extract the ROI (region of interest). Second, the geometrical structure, especially the “T” shape consisting of the aircraft’s swings and fuselage, is extracted based on Hough transform. Third, fine components segmentation based on level set and prior knowledge such as collinearity and symmetry are combined together to identify the components of the target such as the engines and the prow of the aircraft. We get the key parameters of the aircraft from above steps for target recognition and interpretation. The experiments based on the images from miniSAR demonstrate that our method is effective in highresolution SAR images.