he main contribution of this paper is that we proposedan innovative and effective vision system to observe bothposition and spin directly and in real-time mode. With theframe difference-based recognition method, the natural brandon the ball’s surface can be accurately recognized undernormal illumination conditions. By defining the ball’s 3-Dpose in ball’s coordinates, the brand observations of multipantilt vision systems in different frames can be fused and unifiedin one coordinates. Robust and precise spin estimation resultsare gained through the RANSAC-based plane fitting method.Then applying the spin information in a force-based dynamicmodel with the EKF method, a more precise trajectory prediction for a spinning ball can be achieved. Experimentson ping-pong robots verify the effectiveness and accuracy.The methods proposed are very general and can be easilytransformed into other applications such as other ball sportsor even military use.Our future work would focus on the following two aspects.The first is combining our spin vision system with the trajectory deviation-based spin estimating method. This would helpour vision system dealing with the situation when the brandlies at the spin axis and extending the spin observation rangefrom [0, 60 rev/s) to [0, 120 rev/s) (due to Shannon samplingtheorem, the spin vision system can observe only spin speedless than half of its frequency). The second is modeling therebound process between the ball and racquets, which can helpthe robot hitting spinning balls back to a target area on theopponent’s court.