Xingfu Shen

Papers

Total Citations

H-Index

About

Xingfu Shen is a computer vision researcher whose work bridges foundational 3D reconstruction and modern SLAM systems. His most influential contribution, "Surface-Based Structure-from-Motion using Feature Groupings" (2000, 17 citations), introduced a complete pipeline for reconstructing 3D indoor environments from images. The key innovation was a novel matching algorithm that groups features along object boundaries, enabling more robust structure-from-motion in cluttered scenes—a critical advance for early automated modeling. Shen also contributed to the theoretical foundations of geometric computer vision in "Error Propagation from Camera Motion to Epipolar Constraint" (2000, 5 citations), deriving formal relationships between motion parameter perturbations and epipolar errors. More recently, his 2023 work on closed-loop detection for VSLAM (5 citations) addresses a persistent challenge in mobile robotics: improving loop closure accuracy and localization in indoor environments. By proposing an online-updating bag-of-words model for monocular cameras, Shen demonstrates continued relevance in applying geometric principles to real-time robotic perception. His career arc—from foundational structure-from-motion theory to practical SLAM solutions—reflects a sustained focus on making 3D vision systems more accurate and robust for autonomous navigation.