A transformer-based neural network for ignition location prediction from the final wildfire perimeter

Yuming Qiao, Wenyu Jiang, Guofeng Su, Juncai Jiang, Xin Li, Fei Wang

Abstract: Highlights•Traditional ignition traceability methods require heavy human work. So, it's necessary to investigate innovative algorithm.•We implemented a novel transformer-based model to predict wildfire ignition from the final wildfire perimeter.•Our proposed ILNet eliminates the limitations of traditional methods in computational complexity and model accuracy.•The transformer with self-attention mechanism can better extract and represent global spatial features.•The experimental results on real cases show that ILNet has superior performance than traditional CNN models.

External IDs:dblp:journals/envsoft/QiaoJSJLW24