Fire Detection in Infrared Video Surveillance Based on Convolutional Neural Network and SVM

In this paper, a novel algorithm based on convolutional neural network (CNN) and support vector machine (SVM) for fire detection in infrared (IR) video surveillance is proposed. To improve the performance of IR fire detection, we develop a 9-layer convolutional neural network named IRCNN instead of traditional empirically handcrafted methods to extract IR image features. Then, a linear support vector machine is trained with extracted features to achieve fire detection. Our network adopts data augmentation technique and Adam optimization to deal with problems caused by the insufficient dataset, and accelerate the training process. Experimental results show that our method achieved both high precision (98.82%) and high recall (98.58%) on our IR flame dataset and real-time detection on the ordinary infrared surveillance cameras.