| id |
ecaade2022_122 |
| authors |
Kinoshita, Airi, Fukuda, Tomohiro and Yabuki, Nobuyoshi |
| year |
2022 |
| title |
Enhanced Tracking Method with Object Detection for Mixed Reality in Outdoor Large Space |
| source |
Pak, B, Wurzer, G and Stouffs, R (eds.), Co-creating the Future: Inclusion in and through Design - Proceedings of the 40th Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2022) - Volume 2, Ghent, 13-16 September 2022, pp. 457–466 |
| doi |
https://doi.org/10.52842/conf.ecaade.2022.2.457
|
| summary |
Mixed-reality landscape simulation is one of the visual methods used in landscape design studies. A markerless tracking method using image processing has been proposed for properly aligning the real and virtual worlds involved with landscape simulations in large spaces. However, this method is challenging because tracking breaks down if a dynamic object is encountered during the mixed-reality execution. In this study, we integrated deep-learning object detection with natural feature-based tracking, which tracks manually defined feature points (tracking reference points), with the aim of reducing the impact of moving objects such as people and cars on mixed-reality tracking. The prototype system was implemented and tracking was performed on pre-recorded video taken outdoors. Performance was verified in terms of the number of errors associated with tracking the reference points and the accuracy of the mixed-reality display results (camera pose estimation results). Compared to the conventional system, our system was able to reduce the influence of moving objects that cause errors when tracking reference points. The accuracy of the camera pose estimation results was also verified to be improved. This research will contribute to developing mixed-reality simulation systems for large-scale spaces that are accessible to everyone, including users in the architectural field. |
| keywords |
Landscape Visualization, Mixed Reality, Object Detection, Tracking, Deep Learning |
| series |
eCAADe |
| email |
|
| full text |
 file.pdf (1,784,378 bytes) |
| references |
Content-type: text/plain
|
| last changed |
2024/04/22 07:10 |
|
