[1]赵新灿,张燕.基于图形处理器的增强现实自然特征注册算法[J].南京理工大学学报(自然科学版),2011,(04):448-452.
ZHAO Xin-can,ZHANG Yan.Augmented Reality Natural Feature Registration
Based on Graphics Processing Unit[J].Journal of Nanjing University of Science and Technology,2011,(04):448-452.
点击复制
基于图形处理器的增强现实自然特征注册算法
《南京理工大学学报》（自然科学版）[ISSN:1005-9830/CN:32-1397/N]
- 卷:
-
- 期数:
-
2011年04期
- 页码:
-
448-452
- 栏目:
-
- 出版日期:
-
2011-08-31
文章信息/Info
- Title:
-
Augmented Reality Natural Feature Registration
Based on Graphics Processing Unit
- 作者:
-
赵新灿; 张燕
-
郑州大学信息工程学院，河南郑州450001
- Author(s):
-
ZHAO Xin-can; ZHANG Yan
-
College of Information Engineering，Zhengzhou University，Zhengzhou 450001，China
-
- 关键词:
-
增强现实; 三维注册; 通用图形处理器; 自然特征
- Keywords:
-
augmented reality; three-dimensional registration; general-purpose graphics processing unit;
natural feature
- 分类号:
-
TP391. 9
- 摘要:
-
传统的基于中央处理器( CPU) 的计算架构已无法满足增强现实( AR) 三维注册技术实
时运行要求。针对该问题，该文提出采用现代多核异构、大规模并行处理功能的通用图形处理
器( GPGPU) 来加速和优化AR 三维注册算法，研究了在GPU 上实现类似特征提取和匹配等AR
三维注册核心技术的基础理论、方法和实验。结果表明，通过模块划分和优化后的图像特征提
取SIFT 算法、随机采样等，能够充分挖掘AR 三维注册算法的并行运算潜力，对于640* 480 像
素的图像序列，注册算法能够达到15 帧/秒，有效地提高运算实时性。
- Abstract:
-
The traditional computing architecture based on central processing unit ( CPU) can not
meet the real-time requirement of the three-dimensional registration technique for augmented reality
( AR) ． In order to solve the problem，a new idea is proposed here， in which，massively parallel processing
capabilities of general-purpose graphics processing unit( GPGPU) ，with modern multi-core
heterogenety are employed to accelerate and optimize the challenging techniques of the three-dimensional
registration for AR． The fundamental theories，methods and experiments are explored for the
key technologies including the feature extraction and matching． The results show that the
potentialities of AR parallel computing of the 3D registration algorithm such as SIFT image feature
extraction algorithm and random sampling can be further exploited by the methodologies of modularization
and optimization． The registration algorithm can be applied to image sequences with 640×480
pixels at 15 frames /s and the real-time computing ability is improved effectively．
参考文献/References:
［1］ Hall D，Le Gal C，Martin J， et a1． MagicBoard: A contribution
to an intelligent office environment［J］． Robotics
and Autonomous Systems，200l，35 ( 3 ) : 211
－220．
[2] 柳有权，刘学慧，吴恩华． 基于GPU 带有复杂边界
的三维实时流体模拟［J］． 软件学报，2006，17( 3) :
568－578．
[3] 吴仲乐，王遵亮，罗立民． 基于GPU 的快速level set 图
像分割［J］． 中国图象图形学报， 2004，9
( 6) : 679－683．
[4] Comport A I，Marchand E，Prsssigout M，et al． Realtime
markerless tracking for augmented reality: the virtual
visual servoing framework［J］． IEEE Transactions
on Visualization and Computer Graphics，2006，12
( 4) : 615－628．
[5] 袁修国，彭国华，王琳． 基于GPU 的变形SIFT 算子实
时图像配准［J］． 计算机科学， 2011， 38( 3) : 300－303．
[6] Heymann S，Müller K，Smolic A， et al． SIFT implementation
and optimization for general-purpose GPU［A］．
Proceedings of the International Conference in Central
Europe on Computer Graphics，Visualization， and Computer
Vision( WSCG’07) ［C］． Plzen，Czech Republic:
Union Agency-Science Press， 2007: 317－322．
[7] Taylor C J，Kriegman D J． Structure and motion from
line segments in multiple images［J］． IEEE Transactions
on Pattern Analysis and Machine Intelligence，1995，
17( 11) : 1021－1033．
备注/Memo
- 备注/Memo:
-
基金项目: 国家公益性行业科研专项( 200909106) ; 国家“863”计划资助项目( 2007AA040406)
作者简介: 赵新灿( 1972－) 男，博士，讲师，主要研究方向: 增强现实，虚拟维修技术，E-mail: zhaoxincan@126． com。
更新日期/Last Update:
2012-10-23