Visual-based Servo Control method is widely used in small unmanned weapon platforms. We propose a PBVS (Position-based Visual Servo) control method with the aim of obtaining the accurate servo posture and improving the accuracy of the servomechanism platform. Firstly, the target coordinate position is deduced with homogeneous coordinates. Then, the target position is filtered and predicted according to the Kalman filter method, as the delay of visual system and control system while tracking moving target is compensated, later on, the target servo posture is calculated. Finally, the feedforward-feedback control model is applied. The experimental validation results show that the PBVS method based on Kalman filter is more stable than the direct IBVS (Image-based Visual Servo) method. The feedforward link in the control system reduces the average tracking deviation to the target effectively, and the control model of feedback and feedforward can improve the overall control performance of the system, which provides a technical approach for controlling in weapon platforms precisely.