Tracking human objects in a network of surveillance cameras has become an essential requirement in current surveillance systems. Increasing efficiency (response time)of these real-time systems is a challenging task. When searching for a better solution, it was discovered that if software solutions are used it is more effective than using powerful hardware to increase the performance. Many systems use motion detection, feature extraction and appearance matching for re-identification of human objects across cameras. A problem that greatly decreases the efficiency of the system is; features of the object that is being tracked have to be matched with all the objects in all other cameras in the network. This research introduces different strategies that resolve the above mentioned problem of unnecessary feature matching, and enhance the system performance.