Forest Fire Detection Using Combined Architecture of Separable Convolution and Image Processing

Early detection and classification of wildfires using aerial image-based computer vision algorithms like convolution neural networks and image processing techniques have lately gained much attention due to the record-setting wildfire events worldwide. Past studies have demonstrated varying degrees of success in implementing forest fire classification algorithms using variants of well-known sophisticated convolutional neural network architectures, which require extensive computation time for training but demonstrate comparatively high false alarm rates and low predictive power. To accurately detect small-scale forest burns, which typically marks the onset of larger catastrophic events, a combined architecture of separable convolution neural network and digital image processing using thresholding and segmentation is proposed in this paper. The proposed architecture is simple and hence computationally less expensive. Performance evaluation on the test data yielded excellent results in terms of high sensitivity, of about 98.10%, and a low specificity of 87.09%.