WCSE

A Novel Method for Forest Fire Detection Based on Convolutional Neural Network

Abstract— In a forest fire, smoke and fire always appear together. The study of smoke usually uses traditional methods, while deep learning focuses mostly on the features of flame. According to the fact that the smoke is always observed earlier than the flame in the forest fire, this paper presents a deep convolutional neural network model for forest fire detection based on the study of forest fire detecting images. Experimental results show that the deep convolutional neural network for forest fire detection has a higher accuracy than the traditional method in recognition of forest fires by detecting smoke and flame together. In addition, the deep convolutional neural network for forest fire detection combines with preprocessing of ZCA whitening and padding of same size output, which improves the experimental speed and prediction accuracy.

Index Terms— smoke and flame recognition, CNN, forest fire, deep learning

Jialiang Feng, Lihua Liao
School of Computer, South China Normal University, CHINA
Lihua Liao
School of Information Technology in Education, South China Normal University, CHINA

[Download]

Cite: Jialiang Feng, Dingju Zhu, Lihua Liao, "A Novel Method for Forest Fire Detection Based on Convolutional Neural Network," Proceedings of 2018 the 8th International Workshop on Computer Science and Engineering, pp. 264-268, Bangkok, 28-30 June, 2018.