Effective CCTV Patrol Methodology for National Border Defense

Abstract

Due to the low birth rate in Korea, the number of military personnel is decreasing every year. Accordingly, an intelligent surveillance system is being developed by the Korean government to replace human soldiers. The system employs deep learning technology for the detection of enemies and requires sufficient resolution for accurate enemy detection. It is also necessary to zoom in on areas where a targets is present. Therefore, it is necessary to divide an area into detectable pieces and control PTZ(Pan Tilt Zoom) for each region one by one. In this paper, we present a PTZ control methodology for effective enemy detection in this situation. For this purpose, enemy detection scenarios were simulated, and the most suitable method was derived by measuring the detection success rate and average detection time for each proposed method.