Augmented reality (AR) is a technique that allows to seamlessly composite virtual objects or informa-tion into real scene. Compared to traditional text, images and videos, AR is a more effective and intuitive way for information presentation and has wide applications. Simultaneous localization and mapping (SLAM) is a key fundamental technique for augmented reality, which provides the ability of self-localization in an unknown envi-ronment and mapping the 3D environment simultaneously. The localization and mapping enables fusion of virtual objects and real scenes in a geometrically consistent way. In this paper, we describe the basic principles of Visual SLAM, and introduce some state-of-the-art monocular SLAM methods with deep analysis and comparison. Fi-nally, we discuss some research tendency in recent years and make conclusions.