Research on forest and grassland fire detection algorithm based on Himawari-8 geostationary satellite

As one of the crucial processes for maintaining ecosystem balance, natural fires are often characterized by extensive coverage and unpredictability. However, uncontrolled wildfires pose inestimable threats to ecosystems, the atmospheric environment, and human health. Therefore, scientifically predicting and monitoring the temporal and spatial distribution of wildfires is of great significance for fire prevention and control. Based on remote sensing data from Himawari-8, a new-generation geostationary satellite, this study integrates contextual information and dynamic threshold detection methods. The method employs slope deviations of infrared channels to achieve near-real-time detection of fire points. Experimental results demonstrate that the AHI_IGFDA forest fire detection algorithm proposed in this study can extract fire point information efficiently and rapidly. Quantitative comparisons with the FY-3C/VIRR fire detection algorithm and the official Himawari-8 wildfire product (WLF) confirm its outstanding performance: AHI_IGFDA achieves a precision of 0.61, which is significantly superior to WLF (0.18) and FY-3C/VIRR (0.08). Meanwhile, it maintains a low omission rate of 0.36 compared to WLF (0.42) and FY-3C/VIRR (0.84). The optimization of the infrared gradient in this novel fire detection algorithm endows it with higher detection accuracy and lower miss detection rate, and provides new insights and methods for achieving near-real-time fire detection.