Moving Vehicle Detection Based on RPCA Using Multisquint Spaceborne SAR Images

Abstract

In this letter, a moving vehicle detection method is proposed based on spaceborne synthetic aperture radar images, generated by antenna azimuth steering with different azimuth angles and time lags. First, the multisquint geometry and sequential image model are established, and the variation from stationary scatters over sequence is addressed, which interferes the indication of moving signals significantly. Then, the robust principal component analysis (RPCA) is employed in the image domain as the predetection step. Next, with some spatial filters and thresholding estimators, the moving target signature is exploited to further discriminate the residual false detection. Finally, the proposed method is verified by both simulated and real images based on TerraSAR-X data. Results reveal that the clutter suppression can be effectively achieved.