Surveillance Video Summarization System Based On Multiple Object Tracking

With the rapid development of economic society, a lot of intelligent video surveillance systems lead to massive video data. How to efficiently process and store video, and how to improve the utilization efficiency of video. There has been a hot research in the field of computer vision. To this end, this dissertation proposes a video summarization based on multi-object tracking method. In this dissertation, the main work and innovation points are as follows:(1)We propose an approach to solve multi-object tracking based on Generalized Correlation Clustering Problem (GCCP). The task of multi-object tracking can cast as a graph partitioning problem that takes the form of a GCCP. We solve the data association for one object whose all detections are in a sliding temporal window at a time. Multi-object tracking problem can be divided into two phases using hierarchical data association. Firstly, adaptive length tracklets can be obtained from all detections in a sliding temporal window by using the GCCP. Secondly, we treat tracklets as detections, using the similar method described in first phase to acquire trajectories. Experiments show the proposed method can handle occlusion and ID-Switch effectively in complex scenarios. Our Multiple Object Tracking Accuracy (MOTA) and Multiple Object Tracking Precision (MOTP) have higher recognition rate than the-state-of-the-art.(2) A method to get video abstract based on multi-object tracking in real-time is proposed. The method will put the object that appear in different time into the same section of video. We regroup the detected object in the background frame. These objects can keep their original location. Experiments show our summary video can reduce the video storage, video browsing time and improve the utilization ratio of video.(3) We write an APP of video abstract system. The system is based on MFC interface, which uses C++ programming language and relys on OpenCV. The system has a simple interface, which is easy to operate. It can carry on the effective compression to the original video.