Gait identification from invisible shadows
Sensing Technologies for Global Health, Military Medicine, Disaster Response, and Environmental Monitoring II; and Biometric Technology for Human Identification IX
This paper introduces a person identification system that uses as input the shadow images of a walking person, as projected by multiple lights(in this application invisible/infrared lights); the system uses a database of examples of shadows images of a number of people who walk. While it is accepted that personal identification has a higher correct classification rate if views from multiple cameras are used, most systems use only one camera, mainly because (i) Installation in real-world
... real-world environments is easier, less cameras and no need to synchronize cameras, (ii) Computational cost is reduced. In the proposed system, we obtain the advantages of multiple viewpoints with a single camera and additional light sources. More specific, we install multiple infrared lights to project shadows of a subject on the ground and a camera with an infrared transmitting filter mounted in the ceiling inside of a building. Shadow areas, which are projections of one's body on the ground by multiple lights, can be considered as body areas captured from different viewpoints; thus, the proposed system is able to capture multiple projections of the body from a single camera. We explored in other papers the use of sunproduced shadow for identification of people walking freely in the outdoor. In this paper the application scenario is a system installed at the airport in the areas that precedes the immigration checkpoint. Japan already has health monitoring cameras focused on approaching individuals, to determine their health condition; the here described system would also be installed in such a controlled area with restricted walk corridors of walk and controlled lighting. Gait is a remote biometrics and can provide early warning; on another hand it can be used as corroborating evidence in a multi-modal biometrics system. A database of images including shadows for a set of 28 walking people was collected, and the features extracted from shadow areas by affine moment invariants, after which identification of the subject followed. The experiments using the database show the effectiveness of the proposed method and further prove the superiority of using multiple viewpoints compared to a single viewpoint.