摘要
影像匹配是计算机视觉及数字摄影测量的核心问题,基于特征的SIFT匹配算法由于其较强的匹配能力和良好的健壮性成为研究的热点。但SIFT算子的多量性及提取特征点维度较高的特点,直接影响了匹配速度。因此,本文利用GPU强大的并行处理能力,就SIFT特征点匹配在GPU上的设计与实现进行了详细的介绍,并通过对比实验说明利用GPU对SIFT特征点进行匹配所带来的优势。
Image matching is the core problem of computer vision and digital photogrammetry. Because of strong ability of matching and good robustness, SIFT matching algorithm based on feature has become the research hot spot. But the characteristics of large quantity and the high dimension of the extracted feature points direct effect the matching speed of SIFT operator. So In this paper, u- sing the powerful parallel processing capacity of GPU, the design and implement of SIFT feature points matching on the GPU was intro- duced in detail. And the contrast experiment further shows the advantage of the matching of SIFT feature points brought by the GPU.
出处
《测绘与空间地理信息》
2014年第12期14-16,21,共4页
Geomatics & Spatial Information Technology
基金
国家自然科学基金项目(41201482)资助
关键词
SIFT特征点
并行计算
快速匹配
SIFT feature points
GPU
parallel computing
fast matching
