Abstract
In order to protect major facilities from external intruders, a method for establishing and operating an image surveillance system that detects, classifies, and tracks intruders using an image deep learning technology has been applied in various fields. Since the image deep learning performance depends on the number of pixels of the image object and the quality of the image, it is very important to acquire a good quality of images in a large monitoring area. Based on YoLo(You look once), the detection accuracy according to the image size change of the intrusion object was analyzed. A large number of cameras must be used to monitor large areas, but reducing the number of cameras used to reduce budgets has become an important issue. When using a small number of cameras, a blind spot is created as the surveillance direction of the camera must be shifted according to the movement of the intruder. In this paper, the modeling parameters of the camera were analyzed to establish the intruder monitoring area, additionally, in an environment where Geo-fencing is established using a small number of cameras, the number of cameras required according to the parameter settings of the cameras, and the effect on the handover area between intruder surveillance cameras required to track the movement of intruders were analyzed.