基于光流的人体行为识别

Action Recognition Based on the Optical Flow

人体行为识别已成为计算机视觉中的一个研究热点,并且光流法已被应用到各种应用场合。针对教室内学生的站立和坐下的视频,提出了基于光流的人体行为识别算法。首先获取当前帧的活动点集,从而得到活动区域。根据保存帧的信息统计向上光流和向下光流,结合当前人的状态,判断出人的动作。最后进行人的状态的更新。在整个视频处理过程中,该算法重复以上过程,维持了站立人的状态跟踪。实验结果表明,该算法能够识别出站立和坐下的动作,验证了该算法的有效性和鲁棒性。

Recognition of human action has become a hot research topic in computer vision,and optical flow method has been applied to a variety of applications.The action recognition based on optical flow is provided for the video of student standing up and sitting down in the classroom.First,the active area is maintained through the active point set of the current frame.According to the statistical information of optical flows of the saved frames,the upward and downward optical flows are calculated.The hu man action is determined combining the current state of the human.Finally,the state is updated.Throughout the video process ing,the algorithm above process is repeated,maintaining the status tracking of the standing person.Experimental results show that the algorithm is able to identify the action of standing up and sitting down,which verifies the effectiveness and robustness of the algorithm.