Design and Implementation of an Autonomous Surveillance Tower for Intelligent Monitoring Systems

This study presents the design, implementation, and evaluation of a low-cost autonomous surveillance tower intended for real-time perimeter monitoring in resource-constrained environments. The system integrates an ESP32-Cam–based vision module running the lightweight Fast Object Monitoring (FOMO) model, an ESP32-WROOM communication and control unit, a mobile application for remote operation, and a solar-assisted power management architecture. The proposed system achieves real-time human detection with an F1-score of approximately 85%, operating at an average processing speed of 180 ms per frame and sustaining a frame rate of 5.5 fps. A randomized scanning algorithm enhances unpredictability and reduces blind-spot exploitation, while a PID-based tracking mechanism ensures smooth and precise target following under both single and multi-target scenarios. Secure data transmission is achieved through RSA-based encryption and Firebase Realtime Database integration, enabling near real-time video streaming with an average end-to-end latency of 1.2 seconds. Integrated deterrent subsystems including an audio alarm and a low-amperage perimeter defense mechanism demonstrated high effectiveness during controlled tests. The solar-powered energy system ensured uninterrupted autonomous operation with an average consumption of 2.325 W. While the system shows strong performance for its cost and hardware constraints, limitations include reduced accuracy in low-light conditions, material durability concerns for long-term outdoor use, and moderate latency. Overall, the prototype demonstrates that a robust, scalable, and energy-efficient autonomous surveillance solution can be achieved using neural networks and low-power embedded hardware, offering a practical foundation for future military and civilian perimeter-security applications.