A Proposal of Method for Suspicious Person Detection on P2P Security Camera Network System

Recently, the communication among neighbors is now rare, and is become difficult to watch for suspicious person by neighbor residents themselves. Therefore, we proposed a P2P Security Camera Network System that is detects suspicious person by using video data recorded by a camera on neighbor houses. In this paper, we propose four ways to detect suspicious persons based on classification methods using Local Outlier Factor (LOF) for trajectories of moving objects. And, we evaluate proposing methods the effects of useful dimensions and the number of incorrect judgment result on classification of trajectory pattern of moving objects.