摘要
在运动视频检测中,由于存在噪声和遮挡,仅依靠单帧信息来确认目标的位置是很困难的。提出了一种基于轨迹的运动足球检测和跟踪方案。先利用视觉特征来得到每帧图像的候选球,然后使用卡尔曼滤波器进行预测跟踪,生成初始轨迹。通过轨迹选择,确认真实球的轨迹,从而间接地得到球的位置。此外,结合卡尔曼滤波器预测和线性插值,补充视频中漏检的球的位置。实验结果表明,算法具有较高的准确性,精确率达到85%。
Due to much noise and frequent occlusion, it is difficult to detect the sole ball in one frame in soccer broadcast video.In this paper, a trajectory based scheme is proposed to achieve robust football detection and tracking.Firstly, multiple ball candidates are detected through visual features.Then each candidate is tracked by Kalman filter in successive frames, which produces many initial trajectories.After trajectory selection, the true ball trajectories are verified, which indirectly gets the ball position in each frame.Further more, combining Kalman filter prediction and linear interpolation, some missing balls can be made up.The experiment shows that the given scheme has high effectiveness and the accuracy is up to 85%.
出处
《计算机工程与应用》
CSCD
北大核心
2010年第26期148-151,共4页
Computer Engineering and Applications
关键词
检测
跟踪
卡尔曼滤波
轨迹
线性插值
detection
tracking
Kalman filtering
trajectory
linear interpolation
