Pedestrians Tracking in a Camera Network

With the increase of the number of cameras installed across a videosurveillance network, the ability of security staffs to attentively scan all the videofeeds actually decreases. Therefore, the need for an intelligent system that operatesas a tracking system is vital for security personnel to do their jobs well. Trackingpeople as they move through a camera network with non-overlapping field of view isa challenging issue. It requires re-identification of objects when they leaves onefield of view and appears at another scene later. The appearance of the people ischanged significantly among the cameras due to illumination changes, parametersof camera, viewing angle, and deformable geometry of people. Additionally theobservations of people are often broadly separated in time and space and commonproximity techniques can not be used to constrain possible correspondences. In thispaper a new feature was proposed to represent the appearance of people, which iscapable of dealing with the common illumination changes occurring in indoorenvironment. To produce the proposed feature, first the image was transformed toYCbCr color space. Then diagonal vectors of color co-occurrence matrix of eachindividual were extracted and normalized. Experimental results from a realsurveillance scene showed the efficiency of the proposed method.