A View-Planning Approach to 3D Reconstruction

This paper introduces a novel approach for optimizing image capture using View-Planning (VP) to enhance Gaussian splatting for 3D reconstruction of archaeological sites, specifically focusing on Castellaraccio di Monteverdi. Traditional photogrammetry often produces over-smoothed models with artifacts. Our proposed approach leverages VP planning to select optimal viewpoints, ensuring comprehensive image coverage. We integrate this with Gaussian splatting which outputs highly-realistic 3D reconstructions. Initial evaluations on sample datasets demonstrate the potential of VP-enhanced Gaussian splatting to surpass traditional methods in terms of quality. The paper details our approach, discusses the challenges of 3D reconstruction without VP, and outlines our ongoing and future work, including upcoming tests at Castellaraccio di Monteverdi. Our findings aim to contribute to the field of archaeological documentation and analysis, supporting better preservation and understanding of historical sites.