摘要
LED芯片定位是进行芯片检测、划片、扩晶、固晶及判断芯片电气特性、芯片管脚是否达到要求,能否成功分选LED芯片质量的关键一环。针对这一问题,提出了一种图案匹配的定位方法。首先通过获取物体图像的边界特征点,拟合出不依赖于像素灰度的边界形状几何曲线,即图案模板,然后在待检测的图像中搜索相似的几何形状,达到精确定位LED芯片的目的。由于该图案模板与边界有关,与灰度无关,且能旋转、缩放,因此可大大提高定位的鲁棒性和准确性。实验结果表明,其图像定位误差小于1μm,定位速度大于5粒/s,为检测机、分拣机和固晶机等设备的芯片级高精度定位系统开辟了新的方法。
LED chips location is a key link in chips detection, dicing, wafer expansion, die bond and whether the electrical characteristics of the chips and the chips pin can meet the requirements and whether it can sort the LED chips quality successfully. To solve this problem, the pattern matching locating method is presented. Firstly, a series of boundary curves are not dependent on the pixel grid access to the object' s geometry and then finding similar shapes in the image, this technique is not restricted by particular grayscale level in this way, regardless of the object point of the size and shape can accurately find the object changes, this method could improve fundamental work ability. The results of the experiment show that the image locating precision error is less than 1 I^m, locating speed is greater than 5 grains/second, thus providing a new way for chips high-precision locating equipment system such as detection machines, sorters and die bonders.
出处
《半导体技术》
CAS
CSCD
北大核心
2013年第7期556-559,共4页
Semiconductor Technology
基金
福建省科技厅重大基金项目(2011H6022)
福建省教育厅重点项目(JB11161)
关键词
图案匹配
特征点
LED芯片
定位系统
机器视觉
pattern matching
characteristic point
LED chip
locating system
machine vision