In order to improve the stability and veracity of offshore wind turbine installation, it is necessary to perform the heave compensation of the floating crane. But the ship heave motion measured by the inertial system has the characteristics of the random drift and the advanced phase, which seriously affects the real time ability and accuracy of the heave compensation system. So this paper proposes a ship heave motion measurement method based on UKF multi-step observer and BP residual correction. Based on the Stewart wave motion experiment platform, the ship heave motion is simulated, and the inertial measurement system is used to collect heave displacement and acceleration information, then the state space model of the motion is established. Thus, a UKF observer is established by the state space model, and then the state transition matrix is used to perform multi-step observation according to the advanced phase of the heave motion of dynamic detection, to eliminate the random drift and the advanced phase of heave motion. And then, UKF multi-step observer and observation residual are applied to train the BP residual prediction model, and then the prediction residual is used to correct the observation of the multi-step observer online to improve measurement accuracy. The experiments based on Stewart platform show that the proposed method solves the problems of the random drift and the advanced phase of ship heave motion, and the measurement accuracy can be improved to 90%.