A Hybrid Deep Learning-Based Spatiotemporal Fusion Method for Combining Satellite Images with Different Resolutions

Spatiotemporal fusion (STF) is considered a feasible and cost-effective way to deal with the trade-off between spatial temporal resolution of satellite sensors, generate images high resolutions. This achieved by fusing two types images, i.e., fine but rough resolution, resolution. Numerous STF methods have been proposed, however, it still challenge predict both abrupt landcover change, phenological accurately. Meanwhile, robustness radiation differences multi-source crucial for effective application methods. Aiming solve abovementioned problems, in this paper we propose hybrid deep learning-based method (HDLSFM). The formulates framework robust change information minimal input requirements, which nonlinear relative radiometric normalization, superresolution, linear-based are combined address different landcover, prediction. Four comparative experiments using three popular methods, adaptive reflectance model (STARFM), flexible spatiotemporal data (FSDAF), Fit-FC, as benchmarks demonstrated effectiveness HDLSFM predicting change. time interval prediction base dates, ensures its generation fused time-series data.