Joint ISAR Translational Motion Compensation Method via Alternating Optimization and Adaptive Gradient Descent

Accurate translational compensation is crucial for high-resolution inverse synthetic aperture radar imaging. The parameterized optimization methods can directly compensate for the translational motion from radar raw echoes. However, singlecriterion parameterized compensation methods are prone to local optima and may lead to unstable imaging results. To address these limitations, this study proposes a translational compensation method that integrates image entropy and contrast optimization. This method employs a gradient descent strategy that alternates between optimizing entropy and contrast gradients, while dynamically adjusting the learning rate based on objective function feedback. In non-cooperative target imaging, the order of target motion is usually unknown. For this issue, a strategy of gradually increasing the order is introduced to adaptively refine motion compensation. Additionally, to correct residual phase errors that cause imaging results shifting from image center, a centroid-based gradient descent adjustment approach is proposed. Experiments on simulated and real-measured data validate the proposed method's effectiveness under low signal-to-noise ratio and complex motion conditions. Compared to single-criterion optimization methods, our proposed method converges faster, avoids local optima more effectively, and provides better imaging performance.