An approach to the detection of bank robbery acts employing thermal image analysis

A novel approach to the detection of selected security-related events in bank monitoring systems is presented. Thermal camera images are used for the detection of people in difficult lighting conditions. Next, the algorithm analyses movement of objects detected in thermal or standard monitoring cameras using a method evolved from the motion history images algorithm. At the same time, thermal images are analyzed in order to detect regions of concentrated energy. The detection results are combined in order to detect unusual situations such as people gathered in a small area during the assault. Performance of the proposed system was verified employing a set of recordings of a simulated bank robbery, prepared by the authors in a real bank branch site.