Deep Learning Based Forest Fire Classification and Detection in Satellite Images

Abstract

Forest fires are an important threat to humans and other living creatures, with the development of satellite technology it can be constantly monitored and controlled. Presence of smoke in the atmosphere is the indication of forest wildfires. In fire alarm systems, fire detection plays a crucial part in avoiding damages and other fire disasters that lead to social ramifications. Avoiding large scale fire, effective fire detection from visual scenes is important. To improve fire detection accuracy, an effective approach of a convolutional neural network based Inception-v3 based on transfer learning is designed which train the satellite images and classify the datasets into a fire and non-fire images, confusion matrix is generated to specify efficiency of the framework, then extract the fire occurred region in the satellite image using local binary pattern it reduces false detection rates.