激光与光电子学进展, 2020, 57 (24): 241003, 网络出版: 2020-12-02
结合稠密轨迹与视频显著性特征的人体动作识别 下载: 985次
Human-Body Action Recognition Based on Dense Trajectories and Video Saliency
图像处理 动作识别 稠密轨迹 视频显著性 低秩矩阵分解 稀疏编码 image processing action recognition dense trajectories video saliency low-rank matrix decomposition sparse coding
摘要
传统稠密轨迹算法在人体动作识别中取得了较大的成功,但是其在轨迹的形成过程中将动作产生的轨迹和背景运动导致的轨迹进行了相同处理,导致视频表示过于冗余,识别精度受限。为解决这一问题,首先分析背景运动与行为运动模式的差异性,以特征字典的稀疏系数矩阵为基础,利用低秩分解的方法得到稀疏误差矩阵,进一步求解出视频的显著图,然后以显著图作为依据仅在动作相关区域内形成显著性轨迹,并以此表征人体动作。最后基于公开数据集:UCF Sports数据集和YouTube数据集,验证了本文方法的有效性。
Abstract
The traditional dense trajectory algorithm has achieved great success in human-body action recognition. However, the trajectories of the action and background motions are processed equally during algorithm's formation, which leads to redundant video representation and limited recognition accuracy. In this paper, the patterns of the background and behavioral motions are compared, a sparse error matrix is obtained using low-rank matrix decomposition on the basis of the sparse coefficient matrix of the feature dictionary, and a saliency map is solved. The saliency map is then used as the base for representing human-body action in only the action-related areas. The validity of this method is confirmed based on the open datasets UCF Sports and YouTube.
高德勇, 康自兵, 王松, 王阳萍. 结合稠密轨迹与视频显著性特征的人体动作识别[J]. 激光与光电子学进展, 2020, 57(24): 241003. Deyong Gao, Zibing Kang, Song Wang, Yangping Wang. Human-Body Action Recognition Based on Dense Trajectories and Video Saliency[J]. Laser & Optoelectronics Progress, 2020, 57(24): 241003.