The forest fires are threatening phenomenon for the ecological environment, humans, and infrastructure that imbalances the ecosystem environment. There are various types of fire detection systems available commercially and are used globally for fire detection sensor systems. But these fire detection systems are hard to implement in large areas such as forests, as it would be difficult for maintaining the response delay, high cost and other environmental problems. The present research performed forest fire detection algorithm that sharpened filter images which are used for enhancing the edges of the satellite imagesthrough the procedure of preprocessing an original satellite image. The segmented images are moved from RGB regions to YCbCr color space is applied on the basis of color pixel detection for the candidate fire pixels. An Adaptive Neuro-Fuzzy Inference System (ANFIS) is utilized for the segmentation of unnecessary background objects and along with ANFIS the Back Propagation Neural Network (BPNN) is used for the classification of the non-fired images from the fire ones. The results showed that the proposed ANFIS with BPNN method achieves accuracy of 98.96%, which is better when compared with the existing Particle Swarm Optimization (PSO) with Genetic Algorithm (GA) that achieved 91.25 %.

Adaptive Neuro-Fuzzy Inference System, Back Propagation Neural Network, Forest fires, Unsharp filter