Published Paper Details:

Abstract-With the increase in the evolution of technology, the likelihood of occurrence of various natural and man-made calamities have also drastically risen. Forest fires have become a prominent disaster in recent times. Forest fires not only destroy animal life and vegetation but also affect the overall groundwork of a community. Global warming and climate change have resulted in a spike in the forest fire occurrences over the globe. Forest fires are gaining a lot of attention due to the large-scale destruction that they cause worldwide. Pollution is a subsequent product of forest fires and will directly contribute to the overall deteriorating climate crisis. As of now there isn't any standard fire warning system being implemented for early detection of forest fires. Hence the warning message to fire rescue authorities are often long delayed. Various methods such as sensing the fire using fire sensor, UAV (Unmanned Aerial Vehicle), WSN (Wireless Sensor Network) and satellite-based techniques have been experimented on, previously. We propose a Technique where a classification-based algorithm is used for the detection of forest fires and it is implemented in the following stages. Initially the model is trained on CNN (Convolutional Neural Network) classification algorithm. Background subtraction method is applied to the frames extracted from the video, to detect the region that contains movement. Color detection is then performed on the specific frames. The trained model is then utilized to predict the fire and non-fire probabilities.