摘要
为了减少现有的基于特征提取的自动光学检测系统的用户编程时间以及对用户经验的依赖,提出了一种基于统计建模的图像匹配算法.该算法首先对学习训练的焊点图片进行合格或者不合格区分;然后对合格的样本图片进行灰度级别的统计建模,获得一个标准的学习模板;再将待测元件图片经定位操作后与训练好的标准模板进行匹配,通过计算两者像素点灰度值的差值来对待测元件图片进行合格与否的判定.实验结果表明:基于统计建模的图像匹配算法的误报率小于2%,漏报率为0,且可在满足自动光学检测较高检测精度的前提下,大大减少用户对检测程序的编程时间.
In order to reduce the programming time and overcome the experience dependence on the existing automatic optical inspection(AOI) systems based on feature extraction,an image matching algorithm based on statistical modeling is proposed.In this algorithm,first,qualified sample images of solder joints are separated from the unqualified ones in training.Next,a standard learning template image is formed through a gray-level statistical modeling of the qualified sample images.Then,after an alignment,the component image to be tested is matched with the trained template image.Finally,the difference in pixel point gray is calculated and is used to determine whether the testing component image is qualified or not.Experimental results show that the proposed algorithm,with a false alarm rate of less than 2% and a missing report rate of 0,helps to obtain satisfying accuracy of AOI and greatly reduces the programming time of users for inspection.
出处
《华南理工大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2012年第1期64-68,76,共6页
Journal of South China University of Technology(Natural Science Edition)
基金
国家杰出青年基金资助项目(50825504)
NSFC-广东省自然科学联合基金资助项目(U0934004)
广东省高等学校珠江学者岗位计划(2010)资助项目
广东省重大科技专项项目(2009A080204005)
华南理工大学中央高校基本科研业务费专项资金资助项目(2009ZM0073)
关键词
自动光学检测
焊点
统计建模
图像匹配
automatic optical inspection
solder joint
statistical modeling
image matching