An Intelligent Fire Detection System for Ships Using YOLOv8 Deep Learning Architecture

Fire and explosion incidents on ships are becoming increasingly serious and pose substantial threats to maritime safety and security. The consequences cause in extensive damage to property, loss of life, and harm to the marine environment. Meanwhile, traditional methods focus on handling when the fire has formed and spread in the area without the ability to detect and warn early. This study proposes a YOLOv8 deep learning model to detect fire risks at the initial stage in real time, helping to increase accuracy and the ability to analyze impacts quickly. In addition, the Unity 3D environment helps simulate ships with many diverse fire incident scenarios, rich in situations and levels of complexity, enhancing the learning ability of the model that is difficult to test in practice. The diverse training data for the model is optimized for a wide range of modern ship types. The research findings indicate that this approach achieves an accuracy of up to 90% when compared to other methods, significantly reducing the rate of false alarms in challenging conditions. Ultimately, the ship-mountable model improves the reliability and efficiency of future smart ship applications.