Human Tracking with Sparse Distributed Cameras by Spatial-Temporal Likelihood

Abstract

A lot of security cameras are installed in the station and the downtown area in recent years. However blind areas where no camera covers that area are left when a broad territory is watched (sparse distributed cameras). So it is necessary to estimate the correspondence of moving objects taken by different cameras .The appearance sometimes changes drastically by the illumination change The proposed method uses the visual feature and spatial temporal information to estimate the correspondence of the moving objects. The optimum correspondence is obtained by solving Mixed Integer Programming ( MIP). It shows that robust correspondence of moving objects can be estimated under the various environmental changes.