| Augmented reality is a technology which superimposes and fuses computer-generated virtual information into the real environment to enhance users’ perception of reality.With the development of science and technology,augmented reality technology has been widely used in many fields,and has aroused the researchers’ extensive attention.What’s more,augmented reality applications on mobile devices have gradually entered in the majority of users’ view.Simultaneous Localization and Mapping(SLAM)is a key technology in the field of computer vision and robotics.As a three-dimensional registration technology in augmented reality,SLAM realizes tracking and pose estimation of motion cameras,and is a basic and core technology in augmented reality system.Using SLAM tracking and locating to achieve augmented reality has always been the focus of SLAM research.This thesis first studies the mainstream Visual SLAM algorithm,and proposes a fusion scheme of ORB-SLAM and direct method to improve the overall real-time capabilities of the system.Then researches the attitude computation algorithm of IMU,which is used to compute the camera attitude after SLAM tracking lost.Then builds Android augmented reality system,grabs the images of the real scene through the camera in real time,realizes the pose estimation of the camera based on the SLAM algorithm,and finally renders the image and virtual model on the Android device to achieve the effect of augmented reality.The experimental results show that the Android augmented reality system implemented in this thesis works well on the Android platform and can adapt to complex environment,the virtual model can be displayed on the detected horizontal plane in real time,and it can be displayed correctly in the case of pure rotation after SLAM tracking lost.It achieves a good augmented reality effect of displaying 720P images and a maximum display frame rate of 60Hz,and the user can interact with the virtual model,such as zooming and rotating,and have a good visual experience. |