Cluster Tagging: Robust Fiducial Tracking for Smart Environments

Abstract

Fiducial scene markers provide inexpensive vision-based location systems that are of increasing interest to the Pervasive Computing community. Already established in the Augmented Reality (AR) field, markers are cheap to print and straightforward to locate in three dimensions. When used as a component of a smart environment, however, there are issues of obscuration, insufficient camera resolution and limited numbers of unique markers.

This paper looks at the advantages of clustering multiple markers together to gain resilience to these real world problems. It treats the visual channel as an erasure channel and relevant coding schemes are applied to decode data that is distributed across the marker cluster using an algorithm that does not require each tag to be individually numbered. The advantages of clustering are determined to be a resilience to obscuration, more robust position and pose determination, better performance when attached to inconvenient shapes, and an ability to encode more than a database key into the environment. A real world example comparing the positioning capabilities of a cluster of tags with that of a single tag is presented. It is apparent that clustering provides a position estimate that is more robust, without requiring external definition of a co-ordinate frame using a database.