Highly Efficient Implementation of Null Collision Reverse Monte Carlo Method for Infrared Signature Analysis Based on GPU

As a key target characteristic, the analysis of infrared signature is of great significance for target identification, guiding techniques and low observable aerospace design. The numerical study of infrared signature essentially involves solving radiative transfer equation. In this work, null-collision method has been integrated with the reverse Monte Carlo method to improve computing efficiency. A convergence criterion is also proposed to ensure the accuracy. In order to make this method widely available in complex unstructured meshes with high computing efficiency, a GPU accelerated implementation framework is proposed. Hardware ray tracing cores are mainly leveraged. The null collision reverse Monte Carlo method (NC-RMC) is verified against a standard Reverse Monte Carlo method (RMC). And, the influence of absorption coefficient and grid discretization on performance of this method are investigated based on a series of numerical examples. Results shows that the computing efficiency of our implementation framework is weakly sensitive to the refinement level of mesh resolution. Additionally, when the optical thickness in the field is relatively thin and with a refined mesh resolution, the NC-RMC performs better than standard reverse Monte Carlo method. It indicates that the GPU accelerated implementation of NC-RMC is a good choice for infrared signature analysis.