A Multi-temporal Anomaly Analysis Wildfire Detection Method for Transmission Lines

摘要

Fengyun 4A (FY-4A), the first of China’s second-generation geostationary orbiting weather satellites carries a novel optical instrument named Advanced Geosynchronous Radiation Imager (AGRI). An active wildfire detection method based on the AGRI data was developed and tested in this paper. Most of the existing wildfire detection methods were developed on the spectral features of pixels while the temporal characteristics of observed values were ignored. In the proposed method, multi-temporal AGRI data were used to construct time series for each pixel and the fire pixel detection problem is transferred into an anomaly analysis problem. Wildfire pixels can be detected by analyzing characteristics of variation trend value and contextual tests confirmation. To assess the performance of proposed method, the corridor areas within 3 km of electrical transmission lines in China were selected as study areas. Wildfire detection results demonstrate the effectiveness of proposed method with higher sensitivity to detect even small fires and a low wildfire omission error rate.