Research On The Application Of Online Video Summarization Technology In Fog Computing Environment

As security requirements continue to increase and video surveillance hardware costs continue to decline,more and more cameras are connected to video surveillance networks.The rapid growth of video surveillance data provides an application foundation for big data analysis,but it also brings huge challenges for data storage and retrieval.In recent years,the video summary technology that can represent the original video data with a more streamlined and condensed video summary has gradually received extensive attention and research.However,the existing video summary methods often process video files and cannot directly generate a video summary of a real-time video stream.This makes the application of video summarization technology in video surveillance scenarios severely restricted.At the same time,the large amount of video surveillance data transmitted on the network may cause the cloud computing center to fail to analyze and process each real-time video stream in time.Fog computing is an extension of cloud computing.A distributed architecture is used to process data at the edge of the network,which can ease the computing and storage pressure of the cloud computing center.Therefore,research on online video summary technology for real-time video streaming and video monitoring framework based on fog computing has certain application value.This master thesis focuses on the video summary method based on real-time video streaming and its application in the fog computing environment.Its main contents include the following aspects:(1)A video summary method based on real-time video streaming is proposed.Considering that there is a large amount of redundant information in the video surveillance data and the existing video summary technology is difficult to directly generate the video summary of the real-time video stream,a video summary method for the real-time video stream is proposed.In the process of acquiring a real-time video stream,the video data is analyzed and processed in real time according to the movement changes between frames.The video clips and key frames containing motion information in the real-time video stream are used as the dynamic video summary and the static video summary.The generated video summary reduces the redundant information in the real-time video stream,and uses less video frames to record the motion change information in the real-time video stream.(2)A video monitoring framework based on fog computing is proposed.Considering that the cloud computing architecture cannot meet the processing requirements in low-latency scenarios,on the basis of the cloud computing model,a fog computing layer is added as an intermediate layer between the cloud computing layer and the device layer.The fog node in the fog computing layer is responsible for real-time analysis and response to data from the device layer,and the cloud computing layer is responsible for long-term storage and more in-depth analysis of the data processed by the fog node.Compared with a video surveillance framework that only relies on cloud computing for analysis and storage,this framework can effectively reduce network latency and video data transmission volume.(3)Design and implement a set of intelligent video security system.In the video monitoring framework based on fog computing,fog nodes are used to process the real-time video stream generated by the camera and generate corresponding dynamic video digests and static video digests.The cloud server does not need to continuously record real-time video streams,but only needs to store video summaries containing motion information,thereby avoiding waste of storage space.At the same time,users can remotely query and browse video summaries in cloud servers,improving the efficiency and convenience of video analysis.