Structure from Motion Based on Line-Line Intersections

Conventional Structure from Motion (SfM) techniques typically rely on the extraction and matching of image feature points. However, the effectiveness of such methods is often compromised when faced with challenges presented by textureless or cross-modal images. To address these limitations, this research presents a novel SfM approach centered around the utilization of coplanar line segments and their intersections. Our method capitalizes on coplanar line segment matching to establish intersection point correspondences. Given the frequent occurrence of coplanar lines in real-world scenes, our line segment matching technique offers enhanced accuracy and robustness in comparison to point-based features, especially in scenarios involving textureless images. Importantly, our approach doesn’t necessitate specific positional relationships between line segments, such as parallelism or orthogonality. Empirical findings derived from experiments involving textureless and cross-modal images indicate that our proposed method surpasses existing state-of-the-art techniques in terms of both image matching precision and the accuracy of 3D reconstruction.