摘要
根据粘片机晶元的结构特征,提出一种改进的基于小波变换系数投影的算法.该算法将局部有剧烈变化的信号点的位置顺序记为特征点,使用匹配小波系数直方图之间的相似性进行图像匹配,通过改进的多分辨率结构分析,减少了约70%的计算量,并将晶元定位的运算速度提高至10ms左右.该算法不仅可以精准匹配和定位,显著降低系统的总体计算复杂度,也具有更快的晶元搜索速度,可以应用在其他自动化场景中,进一步提高机器性能.
An improved algorithm based on wavelet transform coefficient projection was proposed.A position sequence of local sharp variation points in such signals was recorded as features,and similarity between wavelet parameters histograms was used for image matching.The proposed approach reduces the overhead of computation by 70% over multi resolution data structure algorithm,and the speed of locating wafer achieves within 10ms.The proposed fast algorithm provides not only the accurate matching and locating method,but also a faster searching ability than that of existing algorithms,which can be easily implanted for further performance enhancement in other automation scenario.
出处
《大连海事大学学报》
CAS
CSCD
北大核心
2011年第2期113-116,共4页
Journal of Dalian Maritime University
基金
中央高校基本科研业务费专项资金(2009JC10)
辽宁省教育厅2010年度高等学校科研项目(L2010061)
关键词
快速匹配
小波变换
机器视觉
晶元黏合
半导体
fast matching
wavelet transform
machine vision
wafer bonding
semiconductor.
