虚实遮挡效果的实现是增强现实技术中的重点和难点。增强现实技术在遥操作机器人等不同领域的应用对虚实遮挡处理的实时性、准确性和对不同环境的适应性提出了更高的要求。本文提出一种新的多层次虚实遮挡处理方法，将基于双目立体匹配的三维重建与基于特征点跟踪的三维位姿估算相结合，应用于带有双摄像头的增强现实中。采用多线程方式和关键帧机制，对关键帧进行三维重建，同时对中间帧进行三维跟踪。在等待三维重建完成的过程中利用最近的建模结果来处理虚实遮挡。这种方法无需预先建模，适用于未知和变化的环境，且能满足实时性要求。“,”Handling occlusion relationships between virtual and real objects correctly is a difficult but important issue in the field of augmented reality technique. Applications of real-time augmented reality such as graphical based predictive control of teleoperated robots demand even higher and faster performance of occlusion handling. Adaptability to different surroundings is also essential consideration. This paper proposed a novel approach for multi-layer occlusion handling in real-time augmented reality applications with two cameras, which combines 3D reconstruction based on stereo correspondence and 3D pose estimation based on feature tracking. Two-threaded structure and key frame based approach is used. Information for 3D reconstruction is extracted from pairs of key frames, meanwhile a 3D tracking process is running through inter frames. Models built for real objects are kept in use during the tracking process until new ones are finished. This approach needs no prior knowledge thus is fit for unknown or shape-changing surroundings and is rapid enough for real-time application.