基于Map Reduce的分布式视频处理平台

Distributed Video Processing Platform Based on Map Reduce

计算能力瓶颈限制了复杂视频事件检测算法在实时系统上的应用。为此,设计一种基于Map Reduce模型的分布式视频处理平台,用算子和算子间连接关系描述算法,将算法在时域上切分为并行计算的独立任务。采用普通计算机搭建基于该模型的视频处理集群,运行按模型组织的视频处理算法。实验结果表明,对于处理密集型的视频分析算法,系统处理能力随集群计算机数量的增加呈近似线性增长,能够满足实时处理需求,具有较强的可扩展性。

Most advanced video incident detection algorithms are hard to apply in real-time systems because of its complexity.A distribute video processing platform based on Map Reduce is presented to solve the computation bottleneck,through which video processing algorithm is paralleled on time domain.A distributed video processing architecture is implemented with general-purpose computer.Test experiment proves the flexibility of system’s computation ability.