关键词: 遥感图像融合, 时空反射率融合, 跨分辨率注意力机制, 非参数Bayesian

Abstract: With the rapid development of remote sensing technology,acquiring remote sensing images that simultaneously possess high spatial and high temporal resolution has become a research hotspot.Traditional single optical sensors are limited by strip width and revisit period,making it difficult to meet both requirements simultaneously.The remote sensing image spatio-temporal reflectance fusion technique effectively addresses this issue by combining images with fine spatial resolution but low acquisition frequency and images with coarse spatial resolution but high acquisition frequency.This paper proposes a method based on a dual-layer spatio-temporal fusion framework,which integrates a cross-resolution attention mechanism and a nonparametric Bayesian dynamic dictionary learning mechanism,aiming to generate fused images with both high spatial and high temporal resolution.Experimental results demonstrate that the proposed method exhibits high fusion accuracy and robustness in regions with phenological changes and abrupt land cover changes,and it can better preserve spectral information and spatial details compared to existing methods.

Key words: remote sensing image fusion, spatio-temporal reflectance fusion, cross-resolution attention mechanism, nonparametric Bayesian