A Novel Method for Building Extraction Based on CSAR Images

Circular synthetic aperture radar (CSAR), as a new imaging model, can be used to image the same scene, obtain scattering information from different azimuth angles, and effectively improve the performance of SAR image interpretation. In the CSAR image, the target of the building has not only the strong scattering characteristic, which is the secondary scattering of the light line that is measured in a certain angle, but also the anisotropic scattering characteristics. Meanwhile, the variation of the scattering amplitude in different angles always obeys certain laws as the Angle of observation changes. The method based on the single feature always only describes part of the target feature and cannot extract the characteristics of the target in a comprehensive way. However, the fusion of excessive characteristics can cause the algorithm complexity to rise dramatically, and the algorithm parameter estimation is more complex based on the fusion of multi-feature. Furthermore, the existing algorithm requires the setting of the super parameter manually. In this paper, a novel method for the extraction of the CSAR image building based on multi-statistical characteristics is proposed, which including the dispersion characteristics of the scattering amplitude as well as the anisotropic scattering characteristics. Moreover, the amplitude and the adjacent domain information are considered simultaneously, which obtains a great opportunity for more accurate extraction of building. By experimenting with the actual measured data of the L band, the proposed method is proved to be better than the existing method. To a certain extent, it suppresses the false alarms generated by other ground targets. Additionally, the method exhibits good extraction performance for buildings of various types and different scenes. In summary, the proposed method has the ability to more effectively extract the scattering characteristics of buildings in CSAR images.