Detection of Forest Fire using Convolutional Neural Networks

Abstract

Forest fire is a dangerous condition when an uncontrolled, unexpected fire occurs in forests. It is extremely spontaneous and very difficult to control that damages millions of hectares of land and poses serious dangers not only to the ecosystem but also to humans. Hundreds of fires occur every year due to different reasons: seasonal dry spells, thunderstorms and volcanic ignition. Forest fires pose significant environmental issues, causing economic and environmental destruction and endangering human lives. For several nations a big issue is the occurrence of forest fires coupled with the inability of fire services to contain them effectively. These countries are also developing new strategies for controlling. Timely identification is one essential element to control such a phenomenon. Several classification approaches have been proposed, but there are disadvantages in the proposed models that lead to inefficiency and inability to produce accurate results. A novel Convolution Neural Network algorithm if and when used provides high efficiency, accuracy, and comparatively less data-training stress when compared to the supervised machine learning algorithms that require manual data-training. The results obtained using this technique have been studied and an accuracy of 94.3 percent has been reported.