摘要
传统灰度归一化互相关匹配算法存在匹配时间太长及不适合旋转情况匹配的缺点。提出改进的层叠二分互相关匹配算法,通过将图像多次缩小一半后进行粗匹配,再依次返回到缩小之前的图像上进行小范围的精匹配来减小匹配时间;并提出旋转搜索窗口的方法来使算法适合旋转的匹配;通过使用增量法计算互相关值来进一步减小匹配时间;最后研究了改进的算法在SMD缺陷检测中的应用并进行实验测试,结果表明算法能够快速准确检测元件的位置偏移和角度偏转。
Traditional grey-normalized cross-correlation matching algorithm has two disadvantages：the matching time is too long and does not be suitable for matching in rotation condition.The improved stratified dimidiate cross-correlation matching algorithm is proposed in the paper. In this algorithm,images are rough-matched first after being lessened to half of original size several times,and then back in turn to previous images to make precise matching in a small search range for reducing matching time.The method of rotating search window is proposed in the paper as well to make the algorithm suitable to the matching in rotation condition.Incremental method is used for calculating the cross-correlation value in order to further reduce matching time.At last,the study has been made on applying the algorithm to SMD defect inspection and the experimental testing is carried on too in this regard.The result shows that the algorithm can fast inspect the position offset and the angle deflection of the elements accurately.
出处
《计算机应用与软件》
CSCD
2010年第11期117-120,共4页
Computer Applications and Software
基金
广东省教育部产学研联合项目(2007A09030271)
关键词
归一化互相关
层叠二分
图像匹配
旋转匹配
SMD缺陷检测
Normalized cross-correlation
Stratified-dimidiation
Image matching
Rotation matching
SMD defect inspection
