Super Cyclone Detection and Tracking Using YOLOv8 Algorithm

The severe potential of cyclones presents a serious risk to infrastructure, property, and human life. For disaster planning and prevention strategies, early and accurate cyclone identification is essential. This work presents a new method for super cyclone detection based on the object detection model YOLOv8 (You Only Look Once version 8). Numerical weather prediction (NWP) models are used in the traditional method of cyclone detection. Although NWP models are useful, they have drawbacks, especially when it comes to long-term trajectory prediction. YOLOv8, a deep learning-based object detection model, is utilized to deliver precise and real-time cyclone detection. The proposed method involves the collection of meteorological satellite images and their preprocessing to create a labeled dataset for training. The YOLOv8 model is then fine-tuned on this dataset, enabling it to detect cyclones in these images. The model is very promising for cyclone identification since it can produce predictions without making strict assumptions. We use several metrics, including the confusion matrix, F1-confidence curve, recall confidence curve, precision confidence curve, precision-recall curve, and training and validation loss, to thoroughly evaluate the model's performance. The results of the research work demonstrate the effectiveness of YOLOv8 in super cyclone detection, achieving high accuracy and real-time capabilities. By adopting this deep learning approach, we aim to enhance early warning systems and improve disaster management for super cyclones, ultimately contributing to increased safety and reduced damages caused by these catastrophic weather events. The research work showcases the potential of YOLOv8 as a valuable tool in cyclone detection and tracking, advancing the field of meteorological forecasting and disaster preparedness.