Joint autofocus and registration for video-SAR by using sub-aperture point cloud

Abstract

Due to the existence of antenna phase center (APC) error, drift occurs on the sequence of images acquired by video-SAR back projection (BP) algorithm with respect to a fixed coordinate system. Theoretical analysis shows that the drift is directly related to APC errors. Thus, we propose a novel joint autofocus and registration framework (JAR-PC) for video-SAR in this paper. By estimating the isometric transformation in the sub-aperture SAR image point cloud via the robust point matching network (RPM-Net), the low-order measurement error of the inertial navigation system (INS) is calibrated accurately. Applying the compensated APC trajectory to the BP algorithm, accurate registration and well-focused W-band video-SAR images are obtained. Experimental results show that JAR-PC can easily expand the aperture length from 3 s to 9 s without any visible defocusing, and coherently fuse multiple sub-aperture SAR images into a long-aperture SAR image. By combining the moving-target back projection (m-BP) algorithm with the shadow-tracking technique, JAR-PC can accurately project the SAR image to a geographical coordinate system with moving targets tracked, refocused and well-indicated.