Visible and Infrared Image Fusion Framework for Fire Semantic Segmentation Using U-Net-ResNet50

Forest fires pose a severe threat to several nations throughout the entire word. It frequently leads to important personal, and environmental losses. Because only infrared or visible images can't provide precise data, the fusion VIS/IR images can improve in fire detection. Therefore, combining VIS/IR images includes thermal radiation data with precise texture information. In the paper, first, we evaluate the efficacy of VIS/IR fusion methods using chosen criteria. Second, for fire segmentation, we employ U-Net with pre-trained ResNet50.Finally, results show that fusion stage before the semantic segmentation stage leads to better results compared to visible images only.