Abstract: Accurate ship velocity inversion is fundamental for traffic management and situational awareness. Due to the short integration time and low resolution, ScanSAR mode cannot get the velocity of targets accurately. Sliding spotlight SAR, with its long integration time, captures abundant target motion information, offering significant potential for ship velocity monitoring. However, this means that a ship's velocity vector may change substantially between the sub-apertures, invalidating the constant-velocity assumption of traditional algorithms. Therefore, this study proposes a two-dimensional velocity inversion algorithm for maritime ships, integrating sub-look decomposition and motion compensation. This method models ships as variable-speed targets, by incorporating sub-aperture decomposition, the method imposes physical motion constraints on the phase-refocusing search space, resulting an accurate target velocity in azimuth. Then, considering heading deviations and incorporating a high-accuracy heading estimation algorithm, the two-dimensional velocity of maritime ships is estimated via vector decomposition. Experimental results using GF-3 sliding spotlight SAR data demonstrate that the proposed algorithm achieves an MAE of 0.25 m/s and an MAPE of 7.39% for velocity inversion. The algorithm also demonstrates promising performance in azimuth velocity estimation for anchored ships, indicating its potential for monitoring slow-moving maritime targets