Next-Generation Vision Intelligence for Wildfire Smoke Detection Using YOLOv8 Architecture

Abstract

Wildfires are one of the most destructive natural disasters, causing severe damage to forests, wildlife, human life, and property. Early detection of wildfire smoke is essential to prevent the rapid spread of fire and to enable timely response from disaster management authorities. Conventional fire detection methods such as manual surveillance and basic sensor-based systems often suffer from delayed detection, limited coverage, and high operational costs. This project presents an intelligent wildfire smoke detection system using the advanced deep learning object detection model YOLOv8 (You Only Look Once Version 8). The proposed system is designed to automatically detect smoke from images captured through ground-based cameras and surveillance systems. The YOLOv8 model is trained on a diverse dataset containing wildfire smoke images under different environmental and lighting conditions to improve detection accuracy and robustness. The system architecture consists of image acquisition, preprocessing, model training, smoke detection, and result visualization through a userfriendly Tkinter graphical interface. The trained model identifies smoke regions by generating bounding boxes along with confidence scores, ensuring reliable and real-time detection performance. Experimental results demonstrate that the proposed system achieves high accuracy, improved precision and recall, and reduced false positives compared to traditional image processing methods. The system can be integrated with surveillance cameras and alert mechanisms to provide an efficient early warning solution for forest fire prevention.