为降低晶圆缺陷对半导体制造的影响,在基于改进的多重中值滤波算法的基础上,以差影法为基本原理,采用归一化互相关的模版匹配方法实现晶圆表面缺陷检测。改进的多重中值滤波算法有效实现噪声点与非噪声点的分辨,归一化模版匹配算法对光照具有很好的鲁棒性。对大量的晶粒进行实验,实验结果表明,该方法可有效检测出晶圆表面的缺陷,精度达到15μm左右,所提检测算法在实际的应用中可代替人工,快速、准确地实现晶圆的缺陷检测。
To reduce the influence of wafer defects on semiconductor manufacturing ,based on the improved multiple median filter algorithm ,a method of detecting wafer surface defect was proposed .The image subtraction and the NCC (normalized cross cor‐relation) pattern matching were adopted .The improved multiple median filter distinguished noise and non‐noise points effective , NCC algorithm was robust against change in illumination .Experiences of wafer die verify that the proposed methods can detect wafer surface defect effective ,and the precision reaches 15μm approximately .The defect detection algorithm can replace human work and detect wafer defect quickly and accurately in the actual application .