A Forest Fire Segmentation Method Based on the Fusion of Reflectivity Factor Map and Temporal Difference Features

To leverage weather radar's real‐time capabilities for forest fire monitoring and enhance segmentation accuracy using reflectivity factor maps (RFM), this study proposes a forest fire segmentation method integrating RFM with temporal difference features. This approach improves adaptability to fire area shapes by incorporating temporal differences and enhances segmentation performance by extracting and fusing multi‐scale deep features during Trans‐UNet sampling. It establishes the first end‐to‐end deep learning framework for radar‐based forest fire segmentation. Experiments with real data show the model achieves 85.8% segmentation precision, 82.3% similarity to ground truth, and an inference speed of 19.7 frames per second, demonstrating significant practical value for engineering applications.