A Refinement Reconstruction Method for Indoor Structures Based on 3D Point Cloud Template Matching

Abstract. Indoor 3D reconstruction is a significant research topic in computer vision and computer graphics, focusing on the construction of complete and accurate models of indoor scenes from 3D point cloud data. Traditional data-driven methods often demonstrate poor robustness, low efficiency, and insufficient semantic information when addressing complex indoor environments. To address these challenges, this paper proposes a variable template matching-based method for indoor 3D scene reconstruction, which reframes the complex reconstruction problem as a matching problem. By adjusting and reconstructing library models according to the original instance parameters of the scene, the proposed method facilitates the fine-grained reconstruction of various complex elements within indoor spaces. Utilizing predefined geometric models and contextual constraints, this approach enhances the precision of indoor scene reconstruction, effectively overcoming the limitations associated with traditional data-driven techniques. Extensive experimental validation confirms the effectiveness of the proposed method, demonstrating its ability to alleviate issues such as point cloud noise, data loss, and occlusions, thereby improving both reconstruction accuracy and efficiency. Furthermore, by enriching the reconstructed models with semantic information, this method provides a more comprehensive data foundation for subsequent applications.