Dual-Branch Remote Sensing Spatiotemporal Fusion Network Based on Selection Kernel Mechanism

Popular deep-learning-based spatiotemporal fusion methods for creating high-temporal–high-spatial-resolution images have certain limitations. The reconstructed suffer from insufficient retention of high-frequency information and the model suffers poor robustness, owing to lack training datasets. We propose a dual-branch remote sensing network based on selection kernel mechanism. comprises super-resolution module, feature extraction difference reconstruction module. Convolution adaptive mechanisms are added module improve robustness. upgrades coarse image transition matching fine image; extracts features supplement module; uses structural similarity fine-difference reconstruction. result is obtained by combining with known image. compound loss function used help training. Experiments carried out three datasets five representative algorithms comparison. Subjective objective evaluations validate superiority our proposed method.