An automated real-time direction-sensitive motion detection system

An image processing system is presented that is aimed specifically at helping airport security guards detect passengers entering a concourse through the egress passageway. Motion detection is achieved through grayscale image differencing between reference and current images that have a mosaic substructure. Motion analysis and prediction are performed using predictor templates, and are primarily controlled by a rule-based system that alternates between updating motion activity and motion prediction mosaic images. The detection system is very resistant to false alarms produced by moving shadows and more trivial motions like arm swinging. The system has enough speed and sensitivity to detect small objects, such as a handgun or small package (presumably enclosing a weapon or bomb), thrown across the camera´s field of view. The system also contains enough robustness to detect a human moving the wrong way in a dense crowd of right-way travellers