摘要：This paper presents semantic-based shot event identification. Based on Dynamic Bayesian Network (DBN), the gap between low-lever features and high-lever semantic will be resolve. We apply the mean-shift algorithm and Kalman filter to identify and track the ball and SURF(Speed up Robust Features) to find the basketball hoop. At last the DBN is applied to identify the shot events. Experimental results have shown our proposed method is effective for basketball event detection.
