Neural Networks for Pattern Recognition
Neural Networks for Pattern Recognition
Design of a Component-Based Augmented Reality Framework
ISAR '01 Proceedings of the IEEE and ACM International Symposium on Augmented Reality (ISAR'01)
| Hi-index | 0.01 |
Augmented Reality research currently aims at extending the working range of applications by combining multiple trackers in adjacent areas. The transition between two such devices leads to discontinuities in the trajectory of a tracked object. The result are "jumps" in visual augmentations shown to the user. We present a three step unsupervised learning algorithm that determines the working areas of involved trackers and the area they overlap, and permanently observes the tracked object's position with regard to these areas in order to enable a smooth interpolation within the overlapping area from one tracker's readings to another's. We have tested the algorithm's performance in an experimental setup. The results show that the method is feasible and only adds a negligible overhead to AR systems.