To control the dynamic positioning system of marine surface vessels with unknown model parameters in the real sea environment, and taking into account the limitation of actuators’ saturation and states unavailability etc., a state observing saturation control scheme for dynamic positioning is proposed. First, the ability of the nonlinear extended state observer (ESO) including the unknown saturated control inputs is analyzed. Based on the existing assumptions, a new one is established and the reasonability of the ESO is proved. Then, the linear ESO as the special case is adopted to observe the kinetic states of the ship. Adaptive laws are devised to estimate the damping term of the model and the uncertain gains of actuators, the control input in the form of pitch rate is further derived, and an auxiliary system is introduced to tackle the actuator saturation. The simulation shows that the proposed scheme much more fits engineering reality than previous schemes, while a high precision of positioning and rational convergence speed are promised.