The Design And Implementation Of Distributed Video System For Rapid Identification And Analysis

Video monitoring system has been widely used in smart city,public security,social management,transportation,military defense,and other major industriesin a nation,and is a critical information system related to national livelihood and social stability.With the rapid development and deployment of high-definition video cameras,a lot of large-scale intelligent video monitoring systems such as Sky Net Project emerged.Distributed intelligent video analysis is the main bottleneck that obstructs the widely deployment of video monitoring system.With the explosive growth of video data,distributed video analysis platform and efficient video analysis algorithm have become the core and key components of the video monitoring system,which reflects the availability and intelligence of such system.To meet the basic requirements of video processing for the video monitoring system,based on large data analysis,distributed mass video data processing platform and accurate recognition algorithm are studied in this paper.A distributed video analysis system for rapid recognition is designed and implemented which realizes the efficiency and theintelligence in video analysis.The proposed system includes a Spark-based video analysis and processing platform,and a face detection and recognition subsystem.To address the issue that conventional distributed processing method cannot meet the requirement of real-time intelligent analysis on large quantity of video data efficiently,in this paper,a video analysis processing platform is designed,which combines Spark(a memory-based distributed compute framework)and HDFS(a distributed file system).The platform increases the efficiency of video data processing and effectively supports intelligent analysis on large quantity of video data.Based on the proposed distributed video analysis and processing platform,we design a face detection and recognition sub-system is designed consisting of video acquisition module,video processing module,face detection module,face modeltraining module and face recognition module.IP Cameras in the local area network are searched via the ONVIF protocol and video capture is made,which enables face detection and rapid recognitionthrough preprocessing,recognition,training,etc.The proposed system was tested thoroughly.The results show that the system meets the basic requirements of safe city,social management,etc.