Paper: People Identification Through Ambient Camera Networks Chen, D., Wactlar, H., Bharucha, A.J., International Conference on Multimedia Ambient Intelligence, Media and Sensing (AIMS) Istanbul, Turkey, April 20, 2007 Abstract:   This paper proposes a robust algorithm based on conditional random field (CRF) to improve people identification in an ambient camera network environment. Ambient cameras collect visual signals of subjects from a distance. The quality of these visual signals are often not adequate to train effective classifiers to identify persons in video. We use a CRF to probabilistically propagate the weak identification results across the temporal-spatial space covered by the camera network. No visual based tracking is necessary for our algorithm. Experiments with a 23-camera network demonstrate that the proposed algorithm leverages the advantage of monitoring from multiple cameras and improves the accuracy of person identification. Close