Augmented Reality System Based On Marker

Augmented Reality (AR) is a combination of virtual environment and real environment. It uses the known information in the real world to track and register, then integrates the virtual objects which is generated by computer into the real world, displays the final image by means of display device at last .AR will make us to have a better understanding of our world.AR has its wide application prospect. Many users have PC and webcam. This meets the basic requirement for desk-top AR system. The registration based on markers has a high accuracy and requires no complex track hardware. The mark is easy to track. So the research of AR system based on markers is somewhat practical and significant.The main purpose of this thesis is to implement a simple AR system based on Marker. According to analyzing result of characteristic of the augmented reality system, we develop this AR system with ARToolKit and OpenCV SDK. Tracking by ARToolKit marker, we use finger-tip as interactive means. The main works have three points:1. Through the study of the concept, application of AR, we summarize the three main technologies, and give the analysis of the pros and cons of these technologies, propose that the marker-based registration suits current AR application because of its portable and high real-time performance.2. We describe the principle of marker-based registration and analyze four kinds of AR tools based on marker which are very popular, then select ARToolKit as our research object. We give a simple introduction of the feature and architecture of ARToolKit, make a deep research for ARToolKit modules and registration algorithm. In view of the pose ambiguity, we improve registration module of ARToolKit. Aming at the defect of ARToolKit registration module—pose ambiguity, uses globally convergent pose estimation algorithm and robust pose estimation algorithm from a planar target which was proposed by Gerald Schweighofer to improve the original registration module. The experiment result shows that the improved module could reduce the probability of pose ambiguity and it is more robust than original registration module.3. To develop a simple finger-interactive AR system based on marker, we use ARToolKit for advanced development, and then use OpenGL and Metasequoia modeling software to do graphics rendering .Using the improved ARToolKit algorithm to realize the registration, we use finger-tip to interactive with the virtual menus to control the virtual model. The action was decided by judging whether the finger-tip was in the virtual menu areas through finger-tip detecting in image space. This AR system has a simple hardware and convenient operation. Experimental results demonstrate the system is effective to detect the correct position of finger-tip and it is real-time when the intensity of light is moderate.