Multifeature Domain Fusion Extraction of Buildings From Airborne L-Band CSAR Images Using Subaperture Image Scattering Characteristics

As one of the critical targets in urban areas, the detection and extraction of buildings in synthetic aperture radar (SAR) images are of paramount importance. Distinguishing artificial structures, such as buildings, from natural backgrounds through feature extraction serves as the foundation for subsequent interpretation. Unlike conventional linear synthetic aperture radar (LSAR), Circular synthetic aperture radar (CSAR) enables multi-aspect observation and can acquire target scattering information from various aspects, which facilitates the extraction of ground targets from the CSAR image. Low-frequency SAR systems (e.g., L-band, P-band) present greater challenges for building extraction due to their longer wavelengths and lower resolution. Consequently, low-frequency SAR is predominantly employed for concealed target imaging and detection, with limited applications in man-made feature extraction. To address these limitations, this paper proposes a novel multi-feature domain fusion-based method for building extraction in low-frequency CSAR images. Specifically, the method exploits two distinct characteristics of buildings in CSAR sub-aperture sequences: strong scattering and anisotropic scattering. For strong scattering, a spatial-frequency matrix decomposition-based strong scattering points extraction (SFMD-SSPE) method is proposed, which integrates frequency-domain and spatial-domain information to identify high-intensity scattering pixels across sub-apertures. For anisotropic scattering, a low-rank anisotropic scattering points extraction (LASPE) method is designed, leveraging the statistical properties of the aspect-domain to isolate anisotropic scattering points. Experimental validation on airborne L-band measured data demonstrates that the proposed method achieves superior performance compared to existing CSAR feature extraction approaches. The fusion of multi-domain features effectively enhances the accuracy and robustness of building extraction.