発表論文
2004.02
Walking Person Identification Dealing With Resolution and Appearance Changes
概要
An omnidirectional sensor is able to acquire a 360-degree field of view around a sensor in real time, and can observe azimuth information of persons. By using global azimuth information via an omnidirectional sensor, we can keep tracking the human face while the person walks around the sensor. In this paper, we propose a person identification method from an omnidirectional image sequence. Under an assumption of smooth human motion, we identify the target person from facial database. Also we make a reliable model dealing with changes of resolution and pose of a face while the person walks.