GPGPU-Assisted Subpixel Tracking Method for Fiducial Markers
この論文をさがす
抄録
With an aim to realizing highly accurate position estimation, we propose in this paper a method for efficiently and accurately detecting the 3D positions and poses of traditional fiducial markers with black frames in high-resolution images taken by ordinary web cameras. Our tracking method can be efficiently executed utilizing GPGPU computation, and in order to realize this, we devised a connected-component labeling method suitable for GPGPU execution. In order to improve accuracy, we devised a method for detecting 2D positions of the corners of markers in subpixel accuracy. We implemented our method in Java and OpenCL, and we confirmed that the proposed method provides better detection and measurement accuracy, and recognizing from high-resolution images is beneficial for improving accuracy. We also confirmed that our method is more than two times as fast as the existing method with CPU computation.------------------------------This is a preprint of an article intended for publication Journal ofInformation Processing(JIP). This preprint should not be cited. Thisarticle should be cited as: Journal of Information Processing Vol.22(2014) No.1 (online)DOI http://dx.doi.org/10.2197/ipsjjip.22.19------------------------------
With an aim to realizing highly accurate position estimation, we propose in this paper a method for efficiently and accurately detecting the 3D positions and poses of traditional fiducial markers with black frames in high-resolution images taken by ordinary web cameras. Our tracking method can be efficiently executed utilizing GPGPU computation, and in order to realize this, we devised a connected-component labeling method suitable for GPGPU execution. In order to improve accuracy, we devised a method for detecting 2D positions of the corners of markers in subpixel accuracy. We implemented our method in Java and OpenCL, and we confirmed that the proposed method provides better detection and measurement accuracy, and recognizing from high-resolution images is beneficial for improving accuracy. We also confirmed that our method is more than two times as fast as the existing method with CPU computation.------------------------------This is a preprint of an article intended for publication Journal ofInformation Processing(JIP). This preprint should not be cited. Thisarticle should be cited as: Journal of Information Processing Vol.22(2014) No.1 (online)DOI http://dx.doi.org/10.2197/ipsjjip.22.19------------------------------
収録刊行物
-
- 情報処理学会論文誌
-
情報処理学会論文誌 55 (1), 2014-01-15
- Tweet
詳細情報 詳細情報について
-
- CRID
- 1050282812881111552
-
- NII論文ID
- 110009660271
-
- NII書誌ID
- AN00116647
-
- ISSN
- 18827764
-
- Web Site
- http://id.nii.ac.jp/1001/00098356/
-
- 本文言語コード
- en
-
- 資料種別
- journal article
-
- データソース種別
-
- IRDB
- CiNii Articles