Surveillance Video Abstract Technology Based On Hadoop

As an important means in the field of public security,surveillance cameras are widely used in all aspects of people's life and work.They shoot 24 hours a day without interruption,generating massive monitoring video data,which makes it extremely difficult to store video.When people need to obtain the desired information from the surveillance video,the traditional fast forward and reverse method will cause a great waste of time,and also cause visual fatigue of the viewer,which may cause some information to be judged to be wrong or missing,so the traditional way is completely unable to meet people's needs.Then how to effectively analyze such a large number of monitoring video;How to quickly and conveniently locate and extract the content that people pay attention to;How to reduce the storage space is an important problem that people need to solve.This paper mainly studies the monitoring video summarization technology based on Hadoop,and forms the video summarization through the process of monitoring the moving objects in video through detection,and then picking and putting them into the background model again,which greatly reduces the original video time.And the Hadoop platform is used for distributed storage of video,which solves the storage problem of massive monitoring video data.The main work of this paper includes the following aspects:(1)This paper also focuses on the Hadoop platform,including the HDFS distributed file system,the Map Reduce data processing framework and the cluster resource management tool YARN.This paper completed the construction of Hadoop platform under Windows,and controlled Hadoop and realized the development of Map Reduce task with the help of Eclipse.(2)This paper introduces the main methods of monitoring video abstract technology based on moving objects,and gives the common algorithms of each part,including inter-frame difference method and background subtraction method for moving objects detection;Statistical average method,code method and mixed Gaussian model for background modeling;Bayesian matting,robust matting,Grab Cut and so on are extracted from moving objects,and the advantages and disadvantages of these algorithms are introduced.(3)The monitoring video summary based on moving object is realized,and suitable algorithms such as interframe difference method,mixed Gaussian model,Kalman filtering,Poisson fusion,etc.are used to generate the monitoring video summary,and the operation process of the algorithm is introduced in detail.The bimodal method is added to reduce the number of iterations when the threshold is iteratively selected by the inter-frame difference method.After the Kalman filter is tracked,the Hungarian algorithm is used to match the target,and the summary of the sample video segment is generated.