Video analytics has gradually increased in recent years. The intelligent CCTV cameras in public places, you-tube videos, etc. generate an enormous amount of video data. Generally, video analytics required more time as it contains several processes like encoding, decoding, etc. There are several existing approaches are evolved in improving the efficiency of video analytics but performance delay and loss of data still existing challenges. With our analysis, we strongly state VM migration will be an effective solution to overcome this delay and performance issues. In this paper, we propose NACT based map reducing mechanism (NACT-Map) for processing the real-time streaming videos. The NACT (Novel Awaiting Computation Time) enables the prediction of VM allocation and automatic migration. The scheduling and allocating of the optimal resource are done by task monitor who utilizes the Task manager (TM) system. The NACT based VM migration and MapReduce technique with Hadoop simplifies the process and minimizes the execution time. The splitting of video into chunks of frames speedup the process. Further efficiency is improved by the Map Reduce technique which uses video and its related content for clusters. The performance of our proposed system is executed in the cloudsim with a large dataset contains two real-time videos. Further, the result is compared with the existing methodologies such as distributed video decoding mechanism with extended FFmpeg and VideoRecordReader (VDMFF) (Yoon et al. in Distributed video decoding on Hadoop. IEICE Trans Inf Syst E101-D(1):2933–2941, 2018) and distributed Video Analytics Framework for Intelligent Video Surveillance (SIAT) (Uddin et al. in SIAT: a distributed video analytics framework for intelligent video surveillance. Symmetry 11:911, 2019). The obtained result shows our proposed NACT_Map consumes minimum Task processing time $$({\text{p}}_{{{\text{tix}}}} )$$ ( p tix ) and about 90% of efficiency in overall system performance is increased.

中文翻译：

---

在云服务上使用 NACT-Hadoop/MapReduce 的智能监控视频分析框架

近年来，视频分析逐渐增加。公共场所的智能闭路电视摄像机、Youtube 视频等产生了大量的视频数据。一般来说，视频分析需要更多的时间，因为它包含编码、解码等多个过程。在提高视频分析的效率方面有几种现有的方法正在发展，但性能延迟和数据丢失仍然存在挑战。通过我们的分析，我们强烈声明 VM 迁移将是克服这种延迟和性能问题的有效解决方案。在本文中，我们提出了基于 NACT 的地图缩减机制（NACT-Map）来处理实时流视频。NACT（小说等待计算时间）可以预测 VM 分配和自动迁移。最佳资源的调度和分配由利用任务管理器（TM）系统的任务监视器完成。基于 NACT 的 VM 迁移和 MapReduce 技术与 Hadoop 一起简化了流程并最大限度地缩短了执行时间。将视频拆分为多块帧可加速该过程。Map Reduce 技术进一步提高了效率，该技术将视频及其相关内容用于集群。我们提出的系统的性能在包含两个实时视频的大型数据集的 cloudsim 中执行。此外，将结果与现有方法进行比较，例如具有扩展 FFmpeg 和 VideoRecordReader (VDMFF) 的分布式视频解码机制（Yoon 等人在 Hadoop 上的分布式视频解码中。IEICE Trans Inf Syst E101-D(1):2933–2941 , 2018) 和用于智能视频监控的分布式视频分析框架 (SIAT)（Uddin 等人在 SIAT：用​​于智能视频监控的分布式视频分析框架。Symmetry 11:911, 2019）。获得的结果表明我们提出的 NACT_Map 消耗最少的任务处理时间 $$({\text{p}}_{{{\text{tix}}}} )$$ ( p tix ) 和整个系统大约 90% 的效率性能得到提高。