Abstract: To use Kalman filter method in robot-soccer competition system can eliminate noises contained in vision sensor data,so the position,velocity and acceleration of the moving ball at next several cycles can be predicted accurately.Kalman filter used in linear discrete stochastic system has good convergence and the ability to remove high frequency noises.Performance in contests shows that the proposed method can predict position of the ball more precisely,proving the feasibility and superiority of Kalman filter in the system.