Tracking and pose estimation for computer assisted localization in industrial environments

One of the common needs for many real-time augmented reality (AR) applications is the precise `localization´. Currently at Siemens Corporate Research (SCR), we are developing a real-time system for industrial maintenance assistance. The user is moving within a large industrial site. In order to provide the user with additional information the system needs to locate the user in both real and virtual world. In this scenario, the user is equipped with a mobile computer. The objective is to track and locate the user using a camera attached to the mobile computer. With a set of coded visual markers pre-registered with the global coordinate system, the optical localization could be solved by marker detection, tracking and pose estimation. In this paper, we present the real-time marker detection and pose estimation algorithms used in our mobile localization application. To work in large and complicated industrial environments, our system needs to recover localization information from limited correspondences. We consider two different pose estimation algorithms: the homography based algorithm and the 3-point algorithm. In this paper, we present the results of the numerical experiments comparing these two methods. The experiments are carried out to determine the best approach for our application and to evaluate the accuracy and limitations of the algorithms