Forest fire smog feature extraction based on Pulse-Coupled neural network

A novel algorithm for image-based forest fire smog feature extraction based on Pulse-Coupled neural network (PCNN) is proposed. The PCNN is derived from the phenomena of synchronous pulse burst in mammals' visual cortex. The outputs of PCNN represent unique features of imported images, and has been proved to be invariant to translation, rotation and distortion. In this paper the image from video surveillance monitor is split into three dimensions in RGB color space, then input each dimension to PCNN to extract texture feature, the one-class support vector machine is applied to predict the feature in order to test feature's accuracy. The experimental results show that the algorithm we propose accurately distinguishes smog and non-smog images which outperform both the traditional Euclidean distance algorithms and the algorithms based on grey level co-occurrence matrix (GLCM) we applied in our earlier study. The recognition accuracy is 98% with robustness on our smog image database.