Multi-Ship Target ISAR Imaging Algorithm Based on Multimodal Optimization of Minimum Spatially Weighted Image Entropy

Inverse synthetic aperture radar （ISAR） imaging is a vital technique for maritime surveillance. As maritime targets increase in density， multiple ships with varying motion states often appear simultaneously within the radar’s illumination range. This scenario can result in intersecting or overlapping range profile sequences， leading to mutual interference among echoes and posing significant challenges for echo separation， translational motion compensation， and subsequent imaging. To address these challenges， this paper introduces a multi-ship target ISAR imaging algorithm that employs multimodal optimization of minimum spatially weighted image entropy. Firstly， to address the difficulty in achieving accurate parameter search in multi-target scenarios using the image entropy cost function， we proposed a minimum spatially weighted image entropy optimization criterion. This criterion utilized an anisotropic generalized Gaussian window function to weight imaging results， which effectively reshaped the fitness landscape for parameter optimization and enabled a more robust estimation of the motion parameters of multiple targets. Second， to search for multi-target parameters simultaneously， we designed a hybrid multimodal optimization algorithm based on dynamic population splitting， and achieved synergy between wide-area exploration and local fine search. By simultaneously integrating the fast convergence capability of particle swarm optimization with the strong exploitation ability of white shark optimization， the algorithm simultaneously and precisely estimated the ideal translational motion parameters for multiple targets within a complex parameter space. When combined with the polar format algorithm， this approach facilitated completely processing echo separation， translational motion compensation， and imaging for multiple ship targets， and ultimately produced focused ISAR images of each target. Experimental results using simulated and measured data showed that the proposed algorithm effectively addressed the challenges of multi-target separation and imaging under overlapping range-profile sequences， and achieved high-resolution imaging of multiple targets.