Enhanced Shape-from-Silhouette reconstruction using voxel inverse projection and Marching Cubes Optimization

This study presents an improved Shape-from-Silhouette (SfS) framework for three-dimensional (3D) asteroid reconstruction during the spacecraft approach phases, where imaging resolution is often limited and illumination is uneven. The method integrates voxel inverse projection with the Marching Cubes (MC) algorithm to overcome the efficiency and accuracy limitations of the conventional voxel-based SfS. Morphological operations are used to extract robust contours under varying lighting. Subsequently, voxel inverse projection generates a cumulative value list, followed by MC-based isosurface extraction to produce a continuous 3D point cloud, effectively eliminating staircase artifacts and significantly reducing the computational load. A combined evaluation metric using the Hausdorff distance and peak signal-to-noise ratio is developed to guide threshold selection. Experiments using simulated datasets demonstrate a storage reduction of more than 90% and a runtime reduction of nearly 50% compared with traditional SfS methods while maintaining high geometric accuracy (up to 19.72 dB). These results highlight the potential of the proposed framework for efficient, high-fidelity 3D reconstruction in asteroid proximity operations, supporting navigation, sampling site evaluation, and mission planning for future deep-space exploration.