An Enhanced Algorithm for Active Fire Detection in Croplands Using Landsat-8 OLI Data

Burning biomass exacerbates or directly causes severe air pollution. The traditional active fire detection (AFD) methods are limited by the thresholds of algorithms and spatial resolution remote sensing images, which misclassify some small-scale fires. AFD for burning straw is interfered with highly reflective buildings around urban rural areas, resulting in high commission error (CE). To solve these problems, we developed a multicriteria threshold (SAFD) based on Landsat-8 imagery context croplands. In solving problem CE SAFD algorithm, was LightGBM machine learning method (SAFD-LightGBM), proposed to differentiate fires from sample dataset an optimal feature combining spectral features texture using ReliefF selection method. results revealed that SAFD-LightGBM performed better than method, omission (OE) 13.2% 11.5%, respectively. could effectively reduce interference detection, it has general applicability stability detecting discrete, areas.