When shipboard photoelectric tracking equipment traces target more than 100 km away from it, because of disturbance of barrier, target is often lost from field of view.Under this condition, memory track algorithm shall be adopted to predict location of target at future time to re-find target. Conventional CA and CV models ignore residual error when predicting target, and memory track time is short, which causes insufficient accuracy of target predicted. In consideration of above problems, Kalman target prediction model was put forward to lengthen memory track time. Firstly, this paper derived boat-swing velocity under deck coordinate system from ship-earth coordinate transformation formula, performed feedforward to velocity loop of servo control system to guarantee self-stabilization of optical axis and improve tracking accuracy simultaneously; secondly, CA, CV and Kalman target prediction models were described; finally, 2-dimension position information relationship between memory tracking of 3 kinds of target prediction models and real-time radar guidance location was mainly discussed. Test results show that compared with traditional CA and CV models, target prediction time of proposed method is improved more than an order of magnitudes, that's because Kalman target prediction model is introduced in this paper. The problem in project that tracking accuracy of shipboard photoelectric tracking equipment will be low and memory tracking time will be short when it is affected by boat-swing is solved.