Video Synopsis Based On Multi-target Tracking And Trajectory Combinatorial Optimization

With the rapid development of intelligent video monitoring system, intelligent trans-portation is playing an increasingly important role. Surveillance video processing is facing with the problems of large amount of data and data without structure. People look for-ward to store, browse the information efficiently. Video synopsis is very useful for dealing with the ’big data’ problem.The purpose of this paper is to extract moving objects from video and motion tra-jectory and then combine them in a new way. The existing research methods of object extraction is not accurate. This paper carried out the three aspects of research:(1) the surveillance video target detection and path generation;(2) moving target trajectory combinatorial optimization;(3) framework for generating synopsis in video surveillance. In this paper, the main contributions are as follows:1. This paper has proposed the algorithm of network flow and iterative least square based on Hankel (IHTLS) to track targets. It used the improved network flow algorithm to obtain the initial tracking results, and used IHTLS algorithm to recover lost data. It has improved the initial tracking results, and obtained the target trajectory.2. This paper; has proposed trajectory clustering method for combinatorial opti-mization problems. Energy function was defined to measure the loss of target trajectory regrouping conflicts in space and time.3. This paper has designed a framework of video synopsis algorithm based on multi target tracking. It put the target detection, target tracking and trajectory optimization module integration, and use the’ Poisson Image Editing’techniques to optimize the com-bination of the object and the background picture. The experimental results verify the effectiveness of the proposed method for generating video synopsis.