Toward the optimal spatial resolution ratio for fusion of UAV and Sentinel-2 satellite imageries using metaheuristic optimization

Sentinel-2A/2B twin satellites provide multispectral imagery every 5 days at a medium spatial resolution (10 m for visible and near-infrared bands). In contrast, UAV photogrammetry with low-cost visible-light (RGB) sensors produces ultra-high-resolution orthomosaics but lacks rich spectral information. Fusing these datasets is a solution to enhance the resolution of Sentinel imagery using UAV data. However, fusion is challenging due to large spatial resolution disparities, particularly in Ground Sampling Distance (GSD). By challenging the common practice of sharpening the Sentinel image to match the UAV resolution, we propose a method that fuses the two images at an intermediate resolution level where their information content is comparable. Our approach uses natural target edge analysis and a Genetic-based metaheuristic optimization technique. By minimizing an objective function comprising true GSD or Ground Resolved Distance (GRD) and Mutual Information (MI), we determine the optimal resolution level for fusion. Experimental validation on two UAV and Sentinel datasets, yielded optimal GRD estimates of 2.35 m and 2.03 m, respectively, with Sentinel GRD values of 12.12 m and 12.39 m. The optimal UAV-Sentinel GRD ratios were 0.193 and 0.164. The sharpened Sentinel images showed efficient fusion through subjective and objective quality assessments. Testing the method on 24 UAV-Sentinel datasets from various regions, including America (United States), Europe (Germany, Spain, and Switzerland), and Asia (Iran and Qatar), demonstrated its robustness across different land covers and sensor types. This approach can be applied to any multi-sensor remote sensing image fusion task with significant resolution differences, establishing the meaningful level for fusion.