Behavior prediction of ball carriers in basketball match video

Abstract

This paper proposes a new method of behavior prediction of ball carriers in the basket-ball match video. Due to the low resolution of the head images, the fast motion of sportsmen and the clustered background in the video, this paper proposes the adoption of a covariance descriptor to fuse multiple visual features of the head region, which can be represented as Riemannian Manifolds. Then map the covariance descriptor to the tangent space and complete the head pose classification through the trained multiclass Logitboosts directly in this space for determination of the vision field. According to the distribution of all the sportsmen with-in the range of vision of the ball carrier, this paper predicts his behaviors of shooting, passing and dribbling through sportsmen information based on artificial potential field(APF). At last, this paper testifies the effectiveness of the proposed method by testing on the basketball match videos.

Keywords:

• Covariance descriptor
• Tangent space
• Multiclass LogitBoost
• Artificial potential field