跳水运动视频中的运动目标分割

Automatic Moving Object Segmentation in Diving Video Sequences

针对跳水运动视频中的人体运动目标分割问题,提出一种简单有效的基于颜色特征运动目标检测及分割算法。首帧进行图像预处理,利用色相和亮度两个分量加强颜色差异,然后使用OTSU算法对图像进行阈值分割。结合RGB颜色空间启发式肤色聚类结果确定运动目标所在的连通区域,最终完成首帧运动目标的自动检测。有效地克服了复杂背景变化的影响,能够快速实现跳水运动目标的分割,且对运动对象的快速整体运动有较强的鲁棒性。

In order to segment the moving objects in diving video sequences, all skin color based moving objects segmentation algorithm is proposed in this paper. After increasing difference both hue and luminance elements in first frame, moving objects is detected on the basis of OTSU algorithm and skin color clustering. Experiments demonstrate that our method can overcome the impact duo to complex backgrounds change effectively, extract fastly, and robus to rapid entire movement.