Image Analysis for Detecting Wildfire Using Deep Learning

Abstract

Approximately 80% of the total organisms on Earth rely on trees for sustenance and shelter. Homo sapiens mostly inhabited forested environments throughout their existence. Wildfires have the potential to do harm to forests. Wildfires are an inevitable consequence of the presence of trees. Uncontrolled Wildfires occur in trees above a height of 1.8 meters (6 feet). In recent years, there has been a significant surge in the number of academics who are interested in using computer vision and image processing techniques to detect flames in images. This study examines several deep learning algorithms used in the detection of Wildfires and does a comparative analysis among them. I collect the dataset from online which is analyzed using four different deep learning models: DenseNet201, MobileNetV2, InceptionV3, VGG16. It contains 18,415 images divided into four categories: fire, non-fire, smoke, and fog. The experimental results showed that wildfire predictions have a high level of accuracy, with VGG16 achieving the highest accuracy rate of 97%.