体育视频分析是近年来计算机视觉领域备受关注并且十分具有挑战性的研究方向。提出一种新颖的篮球比赛中持球队员行为预测方法。针对篮球比赛视频中背景复杂,球员移动速度快,头部图像分辨率低的特点,提出运用协方差描述子对持球队员的多种有效头部视觉特征融合,构成黎曼流形,再将其映射到切空间中,用训练好的多类LogitBoost进行持球队员的头部姿态识别,从而确定持球队员的视野范围。根据持球队员视野范围内双方球员分布,提出基于人工势能场的球员信息量,对持球队员投篮、传球、运球3种行为进行预测。最后对实况篮球赛进行测试,验证了这种方法的有效性。

Sports video analysis has received much attention in recent years and is a challenging research direction in the field of computer vision.A novel prediction method for the behavior of ball carriers in basketball matches is proposed in this paper.Aiming at the cluttered background,fast motion of the sportsmen,and the low resolution of the head images in basketball match videos,we propose the adoption of a covariance descriptor to fuse multiple visual features of the head region,which can be represented as Riemannian Manifolds.Then we map the covariance descriptor to the tangent space and complete the head pose classification through the trained multiclass Logitboosts directly in this space for determining the range of vision of the ball carrier.According to the distribution of all the sportsmen within the range of vision of the ball carrier,we predict the behavior of him—shooting,passing, and dribbling,through sportsmen information based on artificial potential field (APF).Finally,the tests on the basketball match videos verify the effectiveness of our method.