Building an advanced invariant real-time human tracking system

Real-time human tracking is very important in surveillance and robot applications. We note that the performance of any human tracking system depends on its accuracy and its ability to deal with various human sizes in a fast way. In this paper, we combined the presented works in [1, 2] to come with new human tracking algorithm that is robust to background and lighting changes and does not require special hardware components. In addition this system can handle various scales of human images. The proposed system uses sum of absolute difference (SAD) with thresholding as has been described in [2] and compares the output with the predefined person pattern using the technique which has been described in [1]. Using the combination between [1,2] approaches will enhance the performance and speed of the tracking system since pattern matching has been performed according to just one pattern. After matching stage, a specific file is created for each tracked person, this file includes image sequences for that person. The proposed system handles shadows removal, lighting changes, and background changes with infinite pattern scales using standard personal computer.