A low-time complexity semi-analytic Monte Carlo radiative transfer model: Application to optical characteristics of complex spatial targets
Peng Gao,Dongxing Tao,Yuan Yuan,Shikui Dong
DOI: https://doi.org/10.1016/j.jocs.2023.101983
IF: 3.817
2023-03-08
Journal of Computational Science
Abstract:The Monte Carlo method (MCM) is one of the most accurate methods for calculating the optical characteristics of spatial targets. However, there are two limitations when using the MCM to build a dataset for a priori phenomenology. One is the multiple scattering effects. The other is photon number configurations in multiple random links (e.g., various irradiance sources, spatiotemporal spectra, complex geometries and structures). Without a reasonable photon number configuration, the dataset's granularity, image quality and computation time will not be compatible. The present work proposes a semi-analytical Monte Carlo radiative transfer model (Semi-MCM), which offers a low order of magnitude time complexity when calculating the scattering properties of the complex spatial target. First, the Semi-MCM was validated theoretically and procedurally. Subsequently, the efficiency of the Semi-MCM is evaluated. The Semi-MCM show a standard double-log-linear relationship between the time complexity and uncertainty. The correlation coefficient is greater than 0.9987. It implies that time complexity can be represented analytically by the single point estimate during calculation. On this basis, the number of photons required for each probabilistic link can be predicted and allocated adaptively to achieve coordination of the probabilistic links. Finally, the distortion from multiple scattering effects is analyzed in terms of the peak signal-to-noise ratio (PSNR). The distortion caused by self-emission is non-negligible, with PSNR in the range of 20 to 40 dB. For the far-field illumination processes of Lambert, the PSNR is in the range of 30 to 40 dB, i.e., the distortion is perceptible but acceptable. The method can provide a basis for generating samples in prior phenomenology and analyzing the influences of multiple scattering effects.
computer science, theory & methods, interdisciplinary applications