A Survey of Augmented Reality Registration Techniques: From Classical Methods to Intelligent Fusion

Registration is a key technology in Augmented Reality (AR), ensuring accurate alignment between virtual content and the real world. This paper reviews the development of 3D registration methods in AR, from traditional vision-based algorithms to modern systems incorporating SLAM, deep learning, and multi-sensor fusion. We categorize and analyze mainstream approaches, summarize their technical features and application scenarios, and compare them in terms of accuracy, robustness, real-time performance, and system complexity. Finally, we discuss current challenges and future trends, including dynamic scene adaptation, semantic understanding, and cloud-edge collaboration, aiming to provide a clear reference for AR research and practical deployment.