Disaster Detection System using YOLO and OpenCV for Real-Time Rescue Operations

Abstract

Natural hazards tend to occur suddenly and may cause serious impacts on property, infrastructure, and human lives. Conventional disaster detection systems that potentially collect disaster information using human reporting mechanisms and sensor-based systems, can cause delays in initiating important response activities. To overcome these existing gaps, we present an Enhanced Disaster Detection System (EDDS) that utilizes YOLOv8 and OpenCV for real-time analysis and monitoring. The EDDS adopts YOLOv8 for visual recognition of disaster indicators, including fire, smoke, flooding, and debris, from live video streams from CCTV or drone cameras. YOLOv8's rapid and precise detection capability aligns perfectly with OpenCV's real time frame processing and enables prompt detection of and alert generation in emergency events. Given the performance characteristics of high accuracy and low latency, our proposed model will improve public safety and expedite readiness for assistance for public safety resources as evidenced by experimental work, and is demonstrated evidence this is a viable option for smart city monitoring and operational disaster management infrastructure.