deFOREST: Fusing Optical and Radar satellite data for Enhanced Sensing of Tree-loss
Researchers have developed a new deforestation detection pipeline called deFOREST that fuses optical and radar satellite data for enhanced sensing. The system constructs anomaly maps from optical data using a discrete Karhunen-Loéve expansion, quantifying anomalies without requiring prior knowledge of data distribution. These optical anomaly maps are then combined with radar data and classified using a Hidden Markov Model. Tested in the Amazon forest with Sentinel-1 and Sentinel-2 data, the deFOREST approach demonstrated superior accuracy compared to existing hybrid methods, particularly in regions with sparse optical data due to cloud cover. AI