With the reductions of hardware cost and the popularity of networks, video surveillance systems usually use multiple video cameras to monitor the environment. Monitoring network consisting of more than one camera can cover a greater area, which is good for the subsequent large-scale video content analysis. In order to achieve this process intelligence, image processing and computer vision methods are needed to achieve multi-camera collaborative processing and analysis. One of the most important research topics is to describe and mine the track of target from different videos, which has great significance in helping police officers handling the case analysis. This thesis implements a multi-camera video object trajectory mining system. On the basis of the extraction of targets from multiple camera video, the system search suspected targets. According to the time information and spatial information of targets, the system will mine the movement trajectory of the targets and present it in the map. In this thesis, the author first gave a brief introduction of related technologies. Then he told the details of the target feature extraction and matching algorithm, which was the basis of track mining algorithm. After that, the author clarified the design of trajectory mining algorithm design. As for the system development, the author first made a comprehensive functional analysis. Then he made the architecture design, module design, interface design and data storage design. And finally, the author described the realization of the system from multiple levels and angles in detail. He also carried out a detailed testing report, which showed that the system meets the functional and performance requirement and it can help improving the work efficiency of police.
修改评论