Complex Cloud-Sea Background Simulation for Space-Based Infrared Payload Digital Twin

The advent of Industry 4.0 has highlighted the requirements for the digitization and intelligent evolution of space-based payloads. To address challenges like limited data samples and simulate infrared images in various scenarios, this study proposes a hybrid data-driven and fractal-driven cloud-sea scenario simulation approach for high-precision infrared images at space-based detection scales. Static cloud-sea scenes are generated using Qilu-2 and New Technology satellite images, while dynamic scenarios are simulated with our iterative fractal dimension optimization algorithm. Next, we propose a high-precision infrared cloud-sea simulation method based on these simulate scenarios. Finally, we validate the confidence of the simulated images through morphological assessment using a 2-D histogram and radiative accuracy evaluation based on Moderate resolution atmospheric transmission (MODTRAN) results. Experimental results confirm the method's accuracy, showing close alignment with on-orbit images. In the 2.7–3.0 μm band, our average radiance is consistent with MODTRAN. Specifically, for reflection angles below 60$^\circ$, the root mean square error between our results and MODTRAN results is about 12.3% in the 3.0–5.0 μm band, and around 3.7% in the 8.0–14.0 μm band. Morphological assessment shows an average error of about 8.3% when compared to on-orbit images. This method allows for generating multiband, multispecies, and multiscale complex cloud-sea scenario images for digital infrared payloads with high flexibility and confidence.