Combined trajectories for action recognition based on saliency detection and motion boundary

作者：

Highlights：

•

• Three sets of trajectories are combined to represent the action in a video.

• More emphasis is put on more informative trajectories in the combination.

• Weighted sparse representation is applied to motion for salient region detection.

• Motion in a short time window is used to avoid missing short-term static area.

摘要

•Three sets of trajectories are combined to represent the action in a video.•More emphasis is put on more informative trajectories in the combination.•Weighted sparse representation is applied to motion for salient region detection.•Motion in a short time window is used to avoid missing short-term static area.

论文关键词：Action recognition,Trajectory-based method,Sparse representation,Saliency detection,Motion boundary

论文评审过程：Received 10 November 2016, Revised 11 May 2017, Accepted 11 May 2017, Available online 13 May 2017, Version of Record 3 June 2017.

论文官网地址：https://doi.org/10.1016/j.image.2017.05.007