Improved ACMMP-3D Reconstruction algorithm based on multi-view geometric consistency constraint priors

This paper explores the challenges of 3D reconstruction in weakly textured regions, emphasizing the limitations of traditional Multi-View Stereo (MVS) algorithms in depth estimation. To address these issues, this study enhances the ACMMP algorithm by incorporating multi-view geometric consistency priors, improving robustness and accuracy. The proposed method uses geometric consistency constraints to generate planar priors and integrates probabilistic models for better depth estimation. Experiments on the ETH3D dataset show a 2.15% improvement in point cloud completeness within a 5 cm threshold, significantly outperforming existing algorithms in weakly textured environments. These findings underscore the enhanced robustness and precision of the proposed method, offering a practical solution for improving 3D reconstruction accuracy, particularly for applications such as medical imaging and cultural heritage preservation.