.png){kind=icon}
返回
Exploiting spatial-temporal context for trajectory based action video retrieval
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-017-4353-2.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Retrieving videos with similar actions is an important task with many applications. Yet it is very challenging due to large variations across different videos. While the state-of-the-art approaches generally utilize the bag-of-visual-words representation with the dense trajectory feature, the spatial-temporal context among trajectories is overlooked. In this paper, we propose to incorporate such information into the descriptor coding and trajectory matching stages of the retrieval pipeline. Specifically, to capture the spatial-temporal correlations among trajectories, we develop a descriptor coding method based on the correlation between spatial-temporal and feature aspects of individual trajectories. To deal with the mis-alignments between dense trajectory segments, we develop an offset-aware distance measure for improved trajectory matching. Our comprehensive experimental results on two popular datasets indicate that the proposed method improves the performance of action video retrieval, especially on more dynamic actions with significant movements and cluttered backgrounds.
Keyword:
Spatial-temporal information
Descriptor coding
Trajectory matching
Bag-of-visual-words
Action video retrieval
AI总结
对已上传原文的论文进行重点信息的提取，主要内容包括：简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。
期刊
IF:
3
论文数:
2.0W
被引数:
3.2W
机构
引用论文
.png){kind=icon}
分享收藏分享收藏
Learning Spatio-Temporal Representations for Action Recognition: A Genetic Programming Approach学习动作识别的时空表示: 一种遗传编程方法
OAAI
OAAI分享收藏
查看更多内容