A fitness-guided adaptive genetic algorithm for near-field neutron coded imaging reconstruction

Neutron coded imaging is a powerful tool for diagnosing the shape and size of the thermonuclear reaction zone in inertial confinement fusion, but achieving high-fidelity reconstruction under near-field conditions remains a challenging, ill-posed inverse problem. In this paper, we propose a novel Fitness-Guided Adaptive Genetic Algorithm (FGAGA) for near-field neutron coded imaging reconstruction. FGAGA employs a fitness-guided adaptive selection operation to dynamically switch between two selection strategies to balance exploration and exploitation, where strategy 1 utilizes parallel processing and strategy 2 utilizes serial processing. The Fitness-Guided Hybrid Crossover Operation and Fitness-Guided Adaptive Mutation Operation (FGAMO) enable the algorithm to accelerate convergence while preserving population diversity. FGAMO further classifies pixels into internal pixels and boundary pixels and designs corresponding neighborhood mutation rules for each category. The algorithm also employs an Isolated Pixel Treatment Operation to suppress the generation of isolated pixels. Reconstructed results demonstrate that FGAGA successfully achieves high-precision continuous grayscale reconstruction of near-field neutron sources under challenging low neutron yields, thereby overcoming the limitation of traditional heuristic algorithms that are restricted to binary reconstruction. Furthermore, the FGAGA shows superior performance compared to established deterministic algorithms. For a near-real neutron radiation source, the FGAGA yields a 2.27 dB gain in PSNR and a 0.12 increase in SSIM over SART with TV.