A Lightweight Multi-View 3D Reconstruction Method for Animal Models Using Specular Reflection with a Single Camera

High-precision 3D reconstruction of animal models plays a critical role in biomedical modeling, ecological monitoring, and digital archiving. Existing approaches such as LiDAR and multi-view stereo (MVS) often rely on expensive hardware setups or complex configurations, limiting their applicability in resource-constrained environments. In this paper, we propose a low-cost specular multi-view 3D reconstruction framework that requires only a single RGB camera and four planar mirrors. This optical setup enables the simultaneous capture of five distinct views of the animal model, significantly reducing hardware costs and simplifying the calibration process. To address geometric distortions and illumination inconsistencies introduced by mirrors, we design a robust image preprocessing pipeline, incorporating adaptive thresholding, morphological denoising, and scale normalization. We further propose a spherical coordinate-based point cloud generation algorithm, combined with multi-view feature consistency verification to enhance reconstruction accuracy. Experiments on various animal models with diverse geometric structures and surface textures demonstrate the effectiveness of our system[1]. Quantitative evaluations using metrics such as Hausdorff distance show that the proposed method achieves an average RMS error of less than 2