With the extensive application of digital technology in the field of cultural heritage, the construction of high-fidelity digital assets has become a core approach for the conservation and inheritance of valuable cultural relics. However, the digitalization of “immovable” relics characterized by large dimensions, fragile materials, and long-term display within climate-controlled showcases—such as the outer coffin of Marquis Yi of Zeng in Hubei Provincial Museum—faces multiple severe challenges, including severely restricted operating space, complex optical conditions, and insufficient material texture. To address these issues, this paper proposes an in-situ non-destructive high-fidelity 3D reconstruction scheme for large and complex relics. First, adhering strictly to the principle of in-situ conservation and targeting the narrow operational conditions in which the showcase glass can only be opened on one side, a dedicated gantry-style expansion acquisition device was independently developed. This modular gantry structure enables standardized image data acquisition with full spatial coverage. Second, at the algorithmic level, a 3D Gaussian Splatting (3DGS) reconstruction framework integrated with multi-view geometric guidance and explicit constraints is proposed. The framework employs Multi-View Stereo (MVS) technology to construct a micro-structural skeleton for scene initialization and introduces single-view smoothness constraints and multi-view geometric consistency verification. This approach effectively overcomes issues such as geometric collapse and artifact accumulation caused by the “high reflectivity and weak texture” of black lacquer surfaces. Experimental results demonstrate that the proposed method not only achieves sub-millimeter geometric reconstruction accuracy, but also faithfully reproduces the unique visual texture of lacquered wooden artifacts. The study provides a new technical paradigm and engineering reference for the in-situ, high-quality digital modeling of large and complex cultural relics.