FIREGUARD AI: A REAL-TIME COMPUTER VISION FRAMEWORK FOR FIRE DETECTION AND AUTOMATED ALERTING USING VIDEO SURVEILLANCE

Fire incidents are highly dynamic and difficult to manage, particularly in environments where smoke, flames, poor visibility, and rapidly changing conditions can make it difficult to identify the exact location and severity of a fire. Conventional fire alarm systems are primarily designed to detect the presence of smoke or excessive heat and provide an early warning. To address this limitation, this paper presents Fireguard AI, a real-time computer vision-based fire monitoring and assistance system designed to provide visual information and contextual details to firefighters during emergencies. FireGuard AI uses live video streams obtained from sources such as webcams, IP cameras, and CCTV systems. The system employs a trained object detection model based on deep learning techniques to identify fire and smoke within video frames. When a potential fire or smoke event is detected, the system highlights the corresponding region in the video and displays the detection confidence, allowing users to visually identify the suspected hazard. The capability of the system to detect fire and smoke in video streams and to give consistent detections from one frame to the next was assessed, and its performance was examined under various conditions such as different smoke densities, lighting, camera position, occlusion, and image quality, all of which can affect the reliability of fire and smoke detection. Overall, FireGuard AI demonstrates the potential of computer vision and artificial intelligence to enhance conventional fire monitoring by providing real-time visual information from existing video surveillance infrastructure. The system integrates fire and smoke detection, confidence-based visualisation, temporal analysis and incident-specific information to provide emergency personnel with relevant information from live video faster and more efficiently.