基于机器视觉的微观表面缺陷检测技术研究
Research on Microscopic Surface Defect Detection Technology Based on Machine Vision

本文研究了一种高速三维微观表面缺陷检测技术，特别针对氮化镓(GaN)硅片在经过薄化研磨后的表面凸起问题。通过发展全域硅片瑕疵侦测法，能够有效避免电性点测时探针撞击凸起导致的损坏。该技术包括三维重建算法和影像校正与细部纹理叠合两部分。利用条纹反射架构，结合相位移及相位展开算法，实现硅片表面的三维形貌重建。通过影像旋转校正和亚像素重建核心图(die map)，准确定位硅片凸点并标记警示区域。该方法具有高效率、高精度和高重复精度的特点，能够满足在线量产设备的检测需求。
This paper studies a high-speed 3D microscopic surface defect detection technology, especially for the surface bulge of gallium nitride (GaN) silicon wafer after thinning and grinding. By developing the global silicon wafer defect detection method, the damage caused by the probe bump during the electrical point measurement can be effectively avoided. The technique consists of a 3D reconstruction algorithm and image correction overlapping with a fine texture. Three-dimensional morphology reconstruction is realized by using the stripe reflection architecture, combined with phase shift and phase expansion algorithm. The core map (die map) is reconstructed by image rotation correction and sub-pixel reconstruction to accurately locate the wafer bumps and mark the warning area. This method has the characteristics of high efficiency, high precision and high repetition accuracy, which can meet the detection requirements of online mass production equipment.