Architecture and Tracking Algorithms for a Distributed Mobile Industrial AR System

Abstract

In Augmented Reality applications, a 3D object is registered with a camera and visual augmentations of the object are rendered into the users field of view with a head mounted display. For correct rendering, the 3D pose of the users view w.r.t. the 3D object must be registered and tracked in real-time, which is a computational intensive task. This contri- bution describes a distributed system that allows to track the 3D camera pose and to render images on a light-weight mobile front end user inter- face system. The front end system is connected by WLAN to a backend server that takes over the computational burden for real-time tracking. We describe the system architecture and the tracking algorithms of our system. Augmented Reality (AR) aims at rendering 3D object information into the users field of view. The virtual elements are merged into the real view by a half- transparent display in front of the users eyes (See-Through-Augmentation) or by mixing the images of a camera with the rendering on an opaque standard display (Video-See-Through, Video-Augmentation). In industrial service augmentation scenarios, 3D information is presented to the service technician into the field of view for hands-free operation. Compositing of virtual objects with real views provides acceptable results only if the 3D position and orientation of the users head w.r.t. the service object is tracked during operation. The tracking allows to adjust the rendering of the virtual elements such that they appear to be fixed on the service object. Tracking is demanding as the technician may move rapidly and look around, loosing sight of the object to be augmented. Time delays in rendering and non-real-time augmentations must be avoided by all means. While real-time tracking is a computationally intensive task, it is also mandatory that mobile AR systems are light-weight, easy to carry, and have low power consumption. These conflicting requirements are solved by your distributed AR system. The system is split into a light-weight mobile front end and a backend computing server which are connected by wireless LAN. The scenario of such a system is sketched in figure 1. This contribution is focused on the design architecture and the tracking al- gorithms for a mobile marker-less vision-based AR system. After a short review on related work we will describe the system architecture and the vision-based tracking algorithms. We close with results on the distributed system.