arXiv · 机器人与计算机视觉· Osman Ali (Manufacturing Metrology Team, University of Nottingham, Nottingham, United Kingdom), Xiangjun Kong (Manufacturing Metrology Team, University of Nottingham, Nottingham, United Kingdom), Tibebe Yalew (Manufacturing Metrology Team, University of Nottingham, Nottingham, United Kingdom), Waiel Elmadih (Taraz Metrology Ltd., Nottingham, United Kingdom), Samanta Piano (Manufacturing Metrology Team, University of Nottingham, Nottingham, United Kingdom)·· 2026-10-07AI 评分10
RACE-FPP:鲁棒的 AI 辅助条纹投影轮廓术表征增强
RACE-FPP: A Robust AI-assisted Characterisation Enhancement for Fringe Projection Profilometry
AI 导读
诺丁汉大学与 Taraz Metrology Ltd.提出 RACE-FPP 框架,将深度学习角点检测融入条纹投影轮廓术(FPP)标准流程以解决成像条件挑战。在混合数据集上,相机重投影误差从 1.237 像素降至 0.259 像素,投影仪误差降低约 50%,显著提升了工业测量鲁棒性。该研究通过全链路误差传播分析验证了系统在模糊及复杂视角下的几何精度优势。
来源:arXiv · 机器人与计算机视觉 · arxiv.org