Design And Application Of Intelligent Video Monitoring System Based On Cloud Outsourcing

With the rapid development of the Network Cloud, Computer Vision technology and network infrastructure construction, a variety of Intelligent Cloud Monitoring system arises at the historic moment. It is able to achieve data backup, multi-terminal login management, intelligent detection and monitoring. And it has many advantages such as data security, monitoring efficiency and initiative. This thesis will study intelligent video surveillance technology based on computer vision technology and the network in-depth, which has certain theory significance and potential application value.The existing monitoring systems are usually stored locally, as after verification tools, they are lack of initiative and weak in data security. In addition, due to each user monitoring is independent, it is difficult to make full use of the monitoring resources and cause serious waste of resources. This thesis proposed a framework design scheme for intelligent video monitoring system based on improved target detection algorithm. This scheme introduces the cloud outsourcing mode. It links each user who has the same monitoring needs, taking high-performance server or server cluster in big data processing and storage as the platform, monitoring unified and management centralized. The cloud outsourcing model, saves the costs of monitoring management, makes full use of monitoring resources, and provides a new reference to the mode of operation of the video surveillance.The system framework mainly includes three parts:distribution platform side, monitoring side and user side. The Real-time video transmission uses RTP/RTCP real-time transport protocol. Each module uses the design method of object-oriented and multi-threaded, which has a good scalability, stability and configurability. Firstly, distribution platform side is the core of the framework. It effectively improves the entire frame work operating efficiency, with the use of asynchronous IO model based on Epoll, message structure based on Google Procotol Buffer, task scheduling mechanism based on boost::asio::io_service, and memory-pool approach. Monitoring side is the backend of the framework, providing real-time video data transmission and video intrusion detection, backup and other functions. The user side is the the front of the framework, providing display of surveillance video real-time or on-demand. The scheme adopts a improved algorithm based on motion history image and frame differential method as moving object detection, which reduces the system’s false detection rate and loss detection rate.In addition, in order to compensate for the inherent performance deficiencies of the single-server mode, the thesis also studied the of the distributed expansion design for distribution platform side. The scheme adopts the master-slave mode of the logic of distribution, using the heartbeat mechanism, load balancing, database synchronization method, and it is capable of supporting fast connections and hot deployment. The scheme provides reference for the design and implementation of large-scale real-time video monitoring and transmission system.