摘要
为了解决传统图像匹配方法在PCB图像匹配过程中准确率低、耗时长的问题,提出一种基于SURF算法结合曲线拟合方法和K-means聚类算法的改进匹配方法。算法如下:首先利用SURF算法提取图像的特征点,并采用最近邻域法对生成特征描述子的特征点粗匹配得到特征点匹配对,然后通过曲线拟合方法滤除部分匹配对,减少匹配耗时,最后采用K-means聚类算法对匹配对聚类分析提取有效的匹配对,完成对特征点的精确匹配。实验结果表明该算法有效剔除了错误的匹配对,提高了PCB图像的匹配精确率,具有较好的稳定性和实时性。
In view of the disadvantages of low accuracy and slow speed in the traditional image matching method in PCB image,this paper proposes an improved image matching method based on the Speeded Up Robust Features（SURF）,which combines the curve fitting method with the K-means clustering algorithm.More specifically,the feature points in the image are extracted by the SURF algorithm and roughly matched by the nearest neighbor method.Then matching pairs of features point are obtained.Also,the partial matching pairs are filtered by the curve fitting method,and the matching time is reduced.Finally,the Kmeans clustering algorithm is used to extract effective matching pairs by clustering analysis,and it completes the exact matching of feature points.Experiments show that the algorithm effectively eliminates the wrong matching pairs,improves the matching accuracy of PCB images,takes less time,and also has good stability and real-time performance.
作者
王平
谢代炎
王杰民
肖国宴
邱芬
WANG Ping;XIE Daiyan;WANG Jiemin;XIAO Guoyan;QIU Fen(a.College of Information Engineering;1 b.College of Science and Technology,Nanchang University,Nanchang 330031,Chin;2.Anyang Test Center for Quality Technology Supervision and Inspection,Anyang Henan 455000,Chin)
出处
《南昌大学学报（理科版）》
CAS
北大核心
2018年第2期184-188,共5页
Journal of Nanchang University(Natural Science)
基金
江西省科技厅科技支撑计划项目(20151BBG70057)
江西省教育厅科学技术资助项目(GJJ14137)
