Hybrid-sampling subpixel-enhanced ray tracing for high-fidelity physical simulation of small targets in infrared complex scenes.

Accurate radiation calculation of subpixel-scale small targets is critical in the physical simulation of infrared scenes. Conventional reverse Monte Carlo ray tracing (RMC-RT) often misses these targets due to random sampling. We propose a hybrid-sampling subpixel-enhanced ray-tracing (HSSE-RT) method to resolve such a problem. By geometry-prior-driven deterministic sampling on the target and stratified random sampling across the background, our method accurately approximates the continuous radiance integral while rigorously ensuring physical energy conservation. Furthermore, we use the strict point spread function (PSF) weighted radiance fusion to calculate the response of pixels. The target intra-pixel offset is explicitly coupled into a position-aware PSF model to accurately capture spatiotemporal energy transitions across consecutive frames during motion. Results demonstrate that our method has high physical fidelity and robustness, providing highly reliable synthetic datasets for applications in infrared target detection systems and optical localization algorithms.