High-Quality Small Fire Samples Generation and Application in Deep Learning-Based Forest Fire Monitoring (opens in new tab)

Early fire detection is critical for disaster mitigation and ecological safety. Geostationary satellites, with their high-temporal-resolution data, offer key support for near-real-time fire monitoring. However, the lack of high-quality fire-sample datasets, especially for small fires, and the underuse of fire-related spectral-spatiotemporal features limit the performance of current algorithms, leading to higher omission errors and reduced effectiveness for early warning. To overcome these lim...