Video surveillance system for community spaces

摘要

This paper proposes a video surveillance system that supports the efficient browsing of surveillance videos captured in community spaces which are daily shared by a certain group of people. Focusing on that only specific persons are likely to appear in community spaces, our system extracts person IDs and their unfamiliarity scores based on their appearance frequencies as the metadata and presents the video clips of persons' appearances on an interface by using the metadata. System evaluation demonstrated that our system enabled users to efficiently browse the surveillance videos by providing person-based search function, the list of appeared persons, and the unfamiliarity score curves.