AntiLoiter

• DOI: 10.1145/2964284.2970927
• Publication Date: 2016-09-29T15:17:32Z
• Authors: Jianquan Liu; Shoji Nishimura; Takuya Araki
• Keywords: Tracking (education); Suspect

Since loitering is a suspicious behavior that often leads to abnormal situations, such as pickpocketing and terrorist attacks, its analysis attracts increasing research attention. In this paper, we present AntiLoiter, practical discovery system for surveillance. AntiLoiter can efficiently discover candidates from longtime videos across multiple cameras. Most of existing systems mainly focus on how detect or identify loiterers by tracking techniques. However, the difficulties tracking-based methods are known their results heavily influenced occlusions, overlaps, shadows. Moreover, need track human appearance continuously. Therefore, they not readily applied cameras due discontinuity criminal loitering. For instance, suspect would probably come same place several times different dates preliminary inspections before attack happens, where discontinuous. To solve problem, abandon method, instead, design novel approach based frequent patterns Our simply yet effectively adopts face recognition techniques discovery. We conducted extensive experiments both synthetic real surveillance evaluate efficiency efficacy our approach. The experimental show find out correctly outperforms method 100 in terms runtime.