基于事件序列匹配的多摄像机视频同步

在多摄像机应用中,对各摄像机的视频进行同步是一个常用的基本步骤。大多数现有方法都依赖于特征点的检测和对应,往往复杂而且难以通用。该文致力于简单快速视频同步方法的研究,通过对常见同步问题的描述和分析,提出了一套基于事件序列匹配的视频同步框架——利用事件序列相关性来帮助寻找两段视频序列中的对应帧,通过全局投票的方法寻求最优帧偏移量。在动作识别和会议建档系统场景中的示例实验中都得到了正确的结果。误差分析表明,该方法的容错能力强,在较大参数范围内都可以得到正确解。该文提出的方法简单易用,具有很好的实用价值。

Event sequence matching based multi-camera video synchronization

Video synchronization is a common issue in multi camera applications with most approaches depending on feature detection and point correspondence which are complicated and difficult to use.This paper provides a simple,fast method for video synchronization based on event sequence matching.This method uses the relationships between event sequences to match frames between two video sequences and seeks global optimum frame offset using statistical methods.Correct results were obtained in tests of action recognition and spot archive scenarios.An error analysis shows that this method is tolerant to errors when the algorithm parameters are properly chosen so the system is convenient to use in practical cases.