Gait Identification Using Invisible Shadows: Robustness to Appearance Changes

摘要: This paper presents a person identification technique that uses information from person's shadow, and is robust to appearance changes caused by variations of clothes carried objects. The invisible lights resulting shadows has advantages undetected sensing. on the ground obtained through illumination multiple can be considered as silhouettes captured virtual cameras placed at light positions. Thus, single camera, e.g. in ceiling, able obtain silhouettes, equivalent multi-camera system. If compared training cases database, bywearing different or carrying a/another bag, then performance gets worse. To deal with this problem, we introduce new shadow-based changes. Firstly, divide each shadow area into several parts, estimate discrimination capability for part based gait features between gallery datasets probe dataset. Next, according estimated capability, adaptively control priorities these parts method. We constructed database variety bags, out successful experiments verify effectiveness proposed technique.