Single view pose estimation of mobile devices in urban environments

摘要: Pose estimation of mobile devices is useful for a wide variety applications, including augmented reality and geo-tagging. Even though most today's cell phones are equipped with sensors such as GPS, accelerometers, gyros, the pose estimated via these often inaccurate, particularly in urban environments. In this paper, we describe an image based localization algorithm estimating Our proposed approach solves homography transformation matrix between phone matching database image, constrained by knowledge change orientation obtained from gyro, 3D information to achieve estimate which improves upon readings GPS compass. We characterize performance dataset Oakland, CA show that query set 92 images, our computed location (yaw) within 10 meters (degrees) 92% (96%) queries compared 31% (26%) (compass) on phone.