Real-Time Human Intrusion Detection in Wireless Sensor Networks Using Yolov7 for Enhanced Perimeter Security

Real time human intrusion detection is another essential part of today's perimeter security, especially for constrained systems such as WSNs. Based on the obtained results, this paper suggests the implementation of a human intrusion detection system with YOLOv7 due to its improved detection performance, power consumption, and real-time performance. The system takes advantage of YOLOv7 for enhanced feature extraction and detection thereby having precision of 94.3 %, recall of 92.8 % and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$F 1$</tex>-score of 93.5. In this work, the authors also propose integrating with WSNs to save energy by 16.4 % compared to similar systems using YOLOv5. A high probability of detecting the movement of objects is achieved, and strong performance is possible at various environmental conditions. The system also increases the alert effectiveness with faster detection to alert time of 90 ms and fewer false & missed detections. These advancements make the proposed approach feasible and efficient for the intended applied real-time perimeter security solution which would be responsive to some of the current unconventional intrusion situations.