Object color propagation in an unregistered distributed video sensor network

In this paper we address the use of object color characteristics as a predominant linking parameter between disjoint video scenes in unregistered distributed video sensor networks. In our approach we proceed by tracking objects in high-resolution video scenes using accumulated frame differencing and morphological techniques. We continue by minimizing shadow effects in object tracking through background detection, in order to obtain more accurate object color signatures. These signatures are then used to link trajectory fragments across disjoint video feeds in a sensor network. We present experimental results to demonstrate the performance of our approach, especially addressing the effect of resolution in tracking, and the role of color signatures as a key linking element when video scene registration is not possible.