A Video Synopsis Method for Object Interactive Preservation Combined with Face Orientation
Qi Tian, Zhixiang Zhu, Chenwu Wang, Pei Wang, Jiangbo Guo, Yuting Wang · 2021 16th International Conference on Intelligent Systems and Knowledge Engineering (ISKE) · 2021
Video synopsis is an effective technology for surveillance video browsing and retrieval. Traditional video synopsis methods focus on reducing collisions among objects and maintaining their chronological order, which may ignore the interaction among the objects and make the interacting tubes lose their original interactivity, resulting in an unreality condensed video. This paper proposes a video synopsis method that preserves the interactive relationship among moving objects. First, a new interactive judgment algorithm is proposed, which uses the height-to-distance ratio and face orientation to determine whether there is an interactive relationship among objects. Second, optimize the correlation energy function, take the interaction relationship as a consideration to avoid destroying the interaction relationship during the video synopsis process. The experimental results show that the proposed video synopsis method can effectively compress the video based on preserving the interaction among the objects.