Development of an Embedded Deep Learning Accelerator for Real-Time Perimeter Security

Abstract

Modern perimeter security systems require intelligent, autonomous, and low-latency monitoring to prevent unauthorized access in restricted environments. This paper presents the development of an embedded deep learning accelerator for real-time perimeter security by integrating computer vision intelligence with FPGA-assisted control. A Raspberry Pi equipped with a camera module captures live video streams and processes them using the YOLO (You Only Look Once) deep learning algorithm for accurate human detection. Upon detecting an unauthorized individual, the Raspberry Pi generates control signals and communicates with an SLG47910 Renesas FPGA through GPIO interfaces. The FPGA acts as a deterministic hardware accelerator to execute response logic such as alert triggering and system control with minimal latency. The hybrid architecture combines software-based intelligence with hardware reliability to achieve fast, scalable, and autonomous surveillance. The proposed system is suitable for military perimeters, restricted industrial zones, and smart border monitoring applications.