Wireless Sensor Network-Driven Human Intrusion Detection Using RT-DETR for Real-Time Perimeter Protection

WSNs are quite important in security concerns where the perimeter is closely monitored to ensure that any intruders are easily detected as they breach the security perimeters. Hence, this work presents an efficient way of developing an RT-DETR (Real-Time Deformable Detection Transformer) that performs human intrusion detection in WSN-based security systems successfully. A model has been proposed which consists of the detection mechanism based on the transformer, well suited for the rapidly changing environments of the security infrastructure for real-time object detection with high accuracy. The experimental results indicate that the proposed method, RT-DETR achieves a detection rate of 97.3%, which is quite better than the other methods such as Faster R-CNN(94.2%), YOLOv5 (95.7%), and SSD-MobileNet <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(91.8 \%)$</tex>. It has false alarm rate of 2.1% in normal lighting conditions and 5.2% in adverse conditions, quite low compare to other system. Further, RT-DETR works with an average of 32 ms inference time at the edge computing environment making it efficient for real-time intrusion detection. Energy efficiency analysis shows that the system overall power consumption is only 120 mW thus making the battery lifespans of WSN nodes to be 28 days longer than that of Faster R-CNN which is <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1 8}$</tex> days and YOLOv5 which is approximately <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{2 2}$</tex> days. These results demonstrate that RT-DETR is beneficial when used to monitor security in real-time security which is critical in faster detection and intervention in WSN-based perimeter protection systems. Further, the advancements in the next years will concentrate on the improved identification of extreme weather conditions during the operation of MG systems, the use of more efficient models of transformers in MG systems, and the integration of AI-based decision-making for the security of MG systems.