Abstract:
In the modern area of warfare and defense, the need for efficient and adaptable surveillance Solutions is Paramount. This thesis leverages a fleet of autonomous and controlled robots equipped with a multitude of sensors, such as H-bridge, ultrasonic sensor, NodeMCU (ESP8266) and camera connected to a Raspberry-pi play a big role. In this thesis, the Yolov8 model as one of the stateof-the-art deep learning models to perform object detection and recognition tasks was applied. The aforementioned robots are deployed in strategic military environments for real –time surveillance and reconnaissance tasks. The core system relies on deep learning(DL) algorithms especially its YOLOV8 model to enable the system to make intelligent decisions. Through this model, the robots continuously learn and adapt their behaviors based on the environment and mission objectives. Robot for military surveillance contains some features such as, communication and coordination. Herein, after capturing the image using camera, the robot processes the image through the YOLOV8 model of deep learning. In addition, it shows and differentiate the detected objects though the identification of name, location, timestamp. It is farther controlled through the control center in real-time while dashboard displays all of data of detected objects with their attributes. This is a real-time decision patrolling specific area, tracking moving targets and reporting anomalies to the dashboard. In other words, by combining IoT and YOLOV8 model, the system offers a scalable, cost-effective and efficiency solution for military surveillance in the camp or in the field enhancing situational awareness and response capabilities. This proposed system provides a potential technology to significantly improve the effectiveness of military operations while reducing risks to realize the capabilities of the innovative approach for military surveillance. The results received indicate that the robot was capable to fully move around and it is able to detect and identify obstacles. The probability of detecting an object as human has a confidence level between 0.5-1.0-while that of other object is about under 0.5.
