Deep Learning-based Border Surveillance System using Thermal Imaging

Border surveillance plays an essential role in border security and the safety of citizens. Borders are monitored 24/7 to maintain law and order within the nation. To control terrorist infiltration and illegal movement of people and goods, it becomes vital to protect the border, even in harsh conditions strictly. Maintaining constant surveillance of border forces uses a lot of manpower and assets; this sometimes even leads to the loss of precision lives of armed forces. Hence, this is the need of the hour for an automated border surveillance framework with high accuracy and low resources requirement. So, this paper proposes a framework to help create a smart border surveillance system based on Machine learning which can be used in any IoT-based device. A study has also been conducted on freezing various layers of the YOLOv5 model. Also, a comparison of the proposed framework is made with existing methods for pedestrian detection. The study shows that lower variants of YOLOv5 have comparable performance as compared to more complex variants for the object detection task. In contrast, much lower GPU and CPU utilization are observed, which enables it to be used in Io devices. Also, the proposed method surpasses the existing methods in terms of mean Average Precision(mAP), Precision, and Recall.