Fire and Smoke Recognition Method for Early Detection of Landfill Fires

The paper is devoted to the development of a method and algorithm for recognizing fire and smoke pixels in images for the purpose of early detection of fires in landfills. A mathematical model for the operation of the proposed method and algorithm has been developed. The YOLOv8 neural network model, which has proven to be quite accurate for such tasks, and a dataset of images for training with the classes "smoke", "fire", "other" and "null" were used for the experiments. The experiments results confirm the performance of the model and method for fire and smoke pixel recognition in the target image.