A Retrieval Method Based on Video Fusion for Surveillance Systems

Digital surveillance systems have already been used in community security widely.However, one main drawback of current surveillance systems is that the user has to search videos from different cameras to retrieve all the video clips of a suspicious object. It is quite inconvenient and time-consuming. In this paper, we propose a video fusion and retrieval method to solve the problem. The method is mainly based on the temporal information, the spatial relationship of cameras, and the object feature to retrieve the related video clips of a target object. A prototype system is implemented for demonstrating the feasibility of the proposed method. The experimental results show that the average successful fusion rate of the proposed method is as high as 82.7%.