Real-Time Fire Detection in Unmanned Ground Vehicles Integrating YoloV5 and AWS IoT

This paper aimed at the development of a remotely controllable 4-wheeled unmanned ground vehicle (UGV) capable of transmitting realtime sensor data and live video feeds over the internet. The central innovation lies in the utilization of an advanced object detection model, to detect fires within the live video feed. By leveraging a Raspberry Pi 4 with 2GB RAM, AWS IoT Core, MQTT, HTTP, and TCP protocols, the UGV achieves seamless data exchange. AWS IoT Core serves as both MQTT broker and subscriber, with the UGV as publisher for sensor readings and subscriber for control commands. The web app acts as the publisher for control commands, generating signals based on user input. The Remote. it software plays a pivotal role in facilitating a cloud-based HTTP service for live video streaming, thereby significantly enhancing global accessibility. Moreover, diverse remote operation methods including keyboard, onscreen, and speech controls, provide flexible means for user interaction. The integration of YOLOv5, Tensorflow, and OpenCV yields a robust fire detection capability, effectively triggering SMTPbased email alerts. This system effectively showcases the seamless fusion of robotics, IoT, web technologies and machine learning, thereby contributing to safety enhancement and inspiring a wave of future advancements.