Design And Implementation Of Surveillance Video Synopsis System

With the rapid development of computer hardware and software technology, a large number of surveillance cameras to be installed in various public places, such as airports, railway stations, hotels, residential, intersection, etc., and responsible for task of monitoring the public safety?Because these surveillance cameras 24 hours a day, 365 days a year of uninterrupted work pattern, and therefore produce a massive surveillance video data.Massive video data also brought bigger challenge, such as artificial check origin video takes a long time and the cost much more than efficient way, finally look for the key events is low efficiency. We can make better use of intelligent monitoring system based on video synopsis technology when video data is so massive, improve data efficiency and precision. In the video surveillance system judged analyze, improve rapid alert, quick browsing and other aspects have a positive meaning.This paper firstly describes the research background and significance, given the present situation of domestic and foreign intelligence surveillance video technology research and development, described the video surveillance system is facing problems and challenges. Then the paper shows the summary of technical theory outlined, technological development, and classification, and focused on the generation algorithm based on the video synopsis. This paper studies and implements the key technology of video synopsis: moving object detection, object tracking, track combination trajectory and generate summary videos fusion core algorithm.In the moving object detection section, this article gives the principles and common algorithms on moving target detection, and focus on the Gaussian mixture background modeling. According to the common characteristics of surveillance video of proposed and implemented based on Gaussian mixture background modeling moving target detection algorithm.In the moving object tracking section, this article describes the principle and moving target tracking algorithm classification algorithm principle is analyzed. And then research and achieve a tracking method which based on color, position features and particle filter multi-target tracking algorithm, in order to achieve the scene for monitoring multiple moving target object trajectories extraction and storage.According to the basic principles of generating a synopsis video, this article proposed object-based video synopsis algorithm which construct the original track and target track energy function mapping the optimal combination regimen trajectory by minimizing energy mapping function, finally through the trajectory with the background image fusion generates a synopsis video.