| 快速互信息匹配及GPU实现 |
| |
| 引用本文： | 田斐,崔玉连.快速互信息匹配及GPU实现[J].电光与控制,2010,17(2). |
| |
| 作者姓名： | 田斐 崔玉连 |
| |
| 作者单位： | 1. 南阳理工学院,电子系,河南,南阳,473004
2. 南阳理工学院,计算机系,河南,南阳,473004 |
| |
| 基金项目： | 河南省重点科技攻关项目资助(0623021600) |
| |
| 摘 要： | 针对模板匹配中速度和精度不能兼顾的问题,提出了一种基于GPU的互信息配准方法。为了算法不至于太复杂而不能在GPU下运行,首先将图像二值化,使得GPU每个线程用到的临时变量减少到4个,然后利用GPU逐像素计算模板和焊缝图像的互信息。将所提算法应用到焊缝定位系统中,通过GPU来执行简化后的互信息计算,基于CUDA的实验表明,所提出的算法可以完成每秒约30帧的图像焊缝定位。
|
| 关 键 词： | 图像处理 模板匹配 互信息 焊缝定位 并行计算 |
Real-Time Mutual Information Matching and Its Implementation on GPU |
| |
| TIAN Fei,CUI Yulian.Real-Time Mutual Information Matching and Its Implementation on GPU[J].Electronics Optics & Control,2010,17(2). |
| |
| Authors: | TIAN Fei CUI Yulian |
| |
| Affiliation: | Nanyang Institute of Science and Technology;a.Department of Electronics;b.Department of Computer Science;Nanyang 473004;China |
| |
| Abstract: | To improve the rate and accuracy of module matching,a new algorithm of real-time module matching was proposed based on mutual-information,which was implemented on Graphics Process Unit(GPU).To reduce the complexity of the algorithm for executing on GPU,image binaryzation was used for reducing the quantity of temporary variables to four.Then GPU was used to calculate the mutual information of module and weldbead image pixel by pixel.The algorithm proposed here was used in a weldbead locating system and GPU w... |
| |
| Keywords: | image processing module matching mutual information weldbead locating parallel computation |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
