Detection Methods Improving Reliability of Automatic Human Tracking System

Much recent research is concerned with overcoming limitations of existing video surveillance systems, particularly for use in automatic human tracking systems. This paper presents detection methods which detect a lost target person during track in automatic human tracking system. The detection methods utilize an algorithm which determines the position of neighbors in a system of video cameras. By utilizing this deployed position and the view distance of video cameras, this algorithm also determines the interrelationship cameras in such a network must have in an automatic human tracking system. The system is enhanced by a video monitoring system utilizing mobile agent technologies. Mobile agents are suitable for distributed processing and parallel processing, since they can monitor their own behavior and run on distributed computers. Multiple mobile agents in the system can track numerous people using information gathered from several neighboring video cameras at the same time. Searching a target person at random is irrational when a system is losing the target; additionally, difficulty of detection can arise if the deployed position and/or the view distance of video cameras vary due to other circumstances. Therefore, a robust computation not influenced by these circumstances is needed, and detection methods utilizing the above algorithm were developed to solve these concerns and to improve reliability of this system by re-detecting the lost target.