165 / 2026-03-31 20:57:12

A 3D Reconstruction Algorithm for ICF Hot-Spot Images under Extremely Sparse Views

ICF hot spot, sparse-view 3D reconstruction, spherical harmonic expansion, 3D Gaussian Splatting, differentiable forward model

激光与粒子束聚变、磁驱动聚变 > 实验与诊断技术

    The three-dimensional morphology of inertial confinement fusion (ICF) hot spots is of great importance for evaluating implosion symmetry and supporting subsequent temperature diagnostics. Owing to geometric constraints of the experimental setup, hot-spot diagnostics usually provide only extremely sparse-view measurement data, making the 3D reconstruction problem severely underdetermined. This work proposes a 3D reconstruction algorithm based on a hybrid model combining spherical harmonic expansion and 3D Gaussian Splatting (3DGS). Spherical harmonics are used to represent the global envelope, low-order asymmetry, and large-scale offset of the hot spot, providing a low-dimensional, smooth, and physically interpretable global prior. 3DGS is used to model local structures such as multi-peak features and residual nonuniform emission, enabling fine representation in continuous space. A unified differentiable forward model is established for joint iterative optimization and reconstruction. Through numerical simulations, the feasibility of the proposed algorithm is verified for reconstructing three-dimensional hot spots from measurement data acquired with one to three orthogonal views, providing an algorithmic basis for subsequent work.