<p>Underwater manipulators, equipped on remotely operated vehicles (ROVs), are widely applied in underwater operations. What challenging is, the base of underwater manipulator is not fixed when the ROV is hovering, which makes the base vulnerable to the influence of water currents. As a result, performing underwater tasks on a hovering ROV becomes burdensome and is even not realizable for human operators with the underwater manipulator end-effector swaying intensely. To compensate for the influence of base swaying for the underwater manipulator end-effector, an active stabilization optimization framework is proposed. Multi-source perception method is proposed to process measurements and perceive a denoised swaying condition, then the trajectory planning method is designed to generate a smooth trajectory with high-order information, which is precisely tracked by the proposed dynamics-based feedforward controller. Consequently, experiments are conducted to demonstrate the effectiveness, and the proposed framework thereby effectively stabilizes the underwater manipulator end-effector, which makes the underwater operation easier as if the ROV has steadily sat and the base is not swaying.</p>