Pedestrian Detection and Tracking in Far Infrared Images

In this paper describes an approach for pedestrian detection in far infrared images. The height of a camera was set up so that people's head might be photoed by the vertical central part of images. In infrared images pedestrians belong to the upper range in the grey-level. Therefore, detects the core (especially face) of a head using the P-tile method. In this approach detects the head area candidate using binarization variable threshold in the search region around core, lowering a threshold. And it makes the search area for detection body and legs. Search area makes below head area candidate. And this method tracks pedestrian after first order prediction. First order prediction uses center of gravity each region and time series.