基于尺度不变特征变换特征点应用于印刷检测的快速匹配算法.pdfVIP

Joumal of Computer Applications ISSN 1 ∞1-9081 2013-06-30 计算机应用，20日， 33( SI): 186 -189 CODEN JYIIDU 文章编号:1001 -9081 (2013)SI -0186-04 基于尺度不变特征变换特征点应用于印刷检测的快速匹配算法 谢文吉\孙晓刚，张亮 (中国科学院成都计算机应用研究所，成都 61∞41) ( *通信作者电子邮箱 xiew叫il0@ mails. gucas. ac. cn) 摘 要:基于机器视觉的质量检测在印刷行业的应用日益重要，而图像匹配算法是其中的关键步骤。针对此问 题提出一种应用于印刷行业的快速高精度匹配算法。首先利用尺度不变特征转换(SI町、)算法精确稳定地提取关键 点，然后通过基于灰皮相关系数的模板匹配算法在定位图像与模板图像的关键点之间找到匹配关系，并结合随机抽 样一致(RANSAC) 方法剔除错误匹配，从而使得整个快速匹配算法高效、稳定、准确。与 SIFT 算法比较，该算法不仅 在时间上快很多，能够满足实时应用的要求，并且极少出现错误匹配。 关键词:图像匹配;模板匹配;尺度不变特征转换算法;印刷质量检测;亚像索 中图分类号:T归91. 413 文献标志码:A Fast matching algorithm in printing quality detection based on scale-invariant feature transform feature points XIE We时i SUN Xiaogang , ZHANG Liang ( Chengdu /nstitωe ofComψter Appl白tion， Chinese Academy of sc岳阳出， Chengdu Sichoon 61∞41 ， China) Abstracl: With rapid development of printing industry recenÙY , the quality detection based on computer vision in the printing industry is becoming increasingly important. Since the image matching algorithm is one of the key steps , this paper pro阳ed a fast high-precision matching algorithm for application of printing industry. Firsùy used the Scale-lnvariant Feature Transform (SIFf) algorithm to extract the key points accurately and stab山ty， then adopted the template matching algorithm , which was based on gray correlation coefficient, to match the relationship between the key points of matching image and template image respectively , furthermore , the RANdom SAmple Consensus (RANSAC) method eliminated false matches. The experimental results show th