Successive Stripe Artifact Removal Based on Robust PCA for Millimeter Wave Automotive Radar Image

This study proposes a stripe artifact removal method based on robust principal component analysis (RPCA) for millimeter wave (MW) automotive radar images. With the development of MW radar detection technology, there is a demand for installing obstacle detectors on vehicular for safety. From this background, the authors developed squint-mode synthetic aperture radar (SAR) with MW (MW-SAR) as a high-resolution imaging technique. For synthesizing radar images, a back-projection algorithm (BPA) is adopted because of its real-time processing nature with high accuracy. However, SAR images obtained with Single Input and Single Output (SISO) systems are prone to be contaminated by a stripe-shaped artifact and can affect to the obstacle detection performance. Thus, to reduce the structured noise, this paper proposes successive RPCA on the assumption that the stripe artifacts and obstacle reflection are low-rank and sparse, respectively. As a solver, the alternating direction method of multipliers (ADMM) is adopted. The main contribution of this work is to initialize the ADMM state by taking account of the similarity of low-rank components between adjacent segments. Through simulations with experimental data, the significance of the proposed method is verified.