Spatiotemporal Fusion Algorithm Based on Multikernal Adaptive Network

Spatiotemporal fusion technology is an effective means to address the challenge of balancing temporal and spatial resolutions for single satellite sensors. There are significant scale differences among various ground objects in remote sensing images; however, existing spatiotemporal fusion models based on convolutional neural networks often employ a fixed receptive field during feature extraction, leading to a lack of dynamic adjustment capability when capturing information at different scales, which in turn affects fusion accuracy. To address this, we propose the Multi-Kernel Adaptive Network (MKAN), specifically designed for spatiotemporal fusion tasks, which achieves adaptive and intelligent extraction, filtering, and fusion of feature information through the innovative Dynamic Visual Processing Module (DVPM) introduced in this paper. In DVPM, multiple parallel branches focus on different feature scales and perform precise fusion through an attention module. This varying attention across branches dynamically adjusts the effective receptive fields of neurons in the fusion layer, enabling precise capture and preservation of critical geophysical features. Furthermore, to enhance fusion efficiency and accuracy, we have innovatively designed the Multi-Dimensional Perceptual Attention Module (MDPA), which integrates channel and spatial domain information enhancement techniques for precise feature identification and optimized pixel-level resource allocation. To comprehensively evaluate the performance advantages of MKAN, we conducted thorough experimental assessments on three datasets and benchmarked it against five spatiotemporal fusion (STF) algorithms. The experimental results demonstrate that MKAN excels in multiple evaluation metrics. Compared to optimal comparison algorithms, MKAN increases structural similarity by an average of 1.4% and reduces global dimensionless error by an average of 4.4%, fully demonstrating the rationality and superiority of its design. Additionally, this chapter explores the effectiveness of the MKAN network structure and assesses the contributions of the DVPM and MDPA modules to the overall algorithm performance through meticulous ablation experiments conducted from various perspectives.