Comparative analysis of YOLOv7 models for use in perimeter protection of military areas

Perimeter protection has crucial importance for the security of forces and deployed military troops in the field. The goal is to protect the base from intruders who can approach on foot or by vehicle. Field wireless sensor networks (WSN) are used to detect intrusions or monitor the secured area’s borders. The camera modules are responsible for capturing images or videos in these sensor networks. In this paper, we will present the YOLOv7 algorithm, which is one of the latest deep learning algorithms for analyzing and object recognition in images and videos collected by WSN. This paper will compare different versions of the YOLOv7 model, using a custom-created data set of images of possible intrusions into given war zones by civilians or civilian vehicles, to select the best model for this type of detection. Testing the algorithms revealed a difference in the final results, and it was found that the YOLOv7x model, which is the most effective and accurate YOLOv7 model, has an average intrusion detection accuracy of 97%.