改进相位相关与聚类思想的Micro-LED芯片显微图像拼接  

Improved phase correlation and clustering ideas for Micro-LED chip microscopic image stitching

作　　者：林坚普 王廷雨 蔡苾芃 张建豪 梅婷 林珊玲 林志贤[1,2,3] 乐建避 吴朝兴 LIN Jianpu;WANG Tingyu;CAI Bipeng;ZHANG Jianhao;MEI Ting;LIN Shanling;LIN Zhixian;LE Jianbi;WU Chaoxing(School of Advanced Manufacturing,Fuzhou University,Quanzhou 362252,China;China Fujian Photoelectric Information Science and Technology Innovation Laboratory,Fuzhou 350116,China;School of Physics and Information Engineering,Fuzhou University,Fuzhou 350116,China)

机构地区：[1]福州大学先进制造学院,福建泉州362252 [2]中国福建光电信息科学与技术创新实验室,福建福州350116 [3]福州大学物理与信息工程学院,福建福州350116

出　　处：《光学精密工程》2025年第4期641-652,共12页Optics and Precision Engineering

基　　金：国家自然科学基金青年基金(No.62101132);福建省自然科学基金资助(No.2024J010016)。

摘　　要：针对现有Micro-LED芯片的显微图像由于存在大量相似且整齐排列的LED而导致算法拼接速度慢及准确率低的问题,本文提出了一种基于聚类思想的Micro-LED芯片显微图像拼接方法。首先使用最大类间方差法对图像进行预处理,随后使用自适应面积拓展相位相关算法对相邻两幅图像的重叠区域进行峰值计算以求得初步配准的偏移量,接着使用图像分块联合差异值哈希算法的优化配准策略得到优化后的偏移量,最后引入基于密度的噪声应用空间聚类对可能存在的错误偏移量进行自动筛出和处理。实验结果表明本文算法平均耗时为64 ms,满足拼接实时性的需求,拼接正确率达到99.4%,配准精度控制在1个pixel之内,融合处理后在主观上可以达到完美的拼接效果,为错误偏移量导致的整体错位问题也提供了新的解决思路。本文算法有效解决了晶圆级Micro-LED芯片检测中显微图像拼接的关键环节,并可推广应用于具有重复特征且高精度要求的其他机器视觉自动化领域,具有较高的实际工程应用价值。Aiming at the problem of slow algorithmic stitching and low accuracy due to the existence of a large number of similar and neatly arranged LEDs in the existing microscopic images of Micro-LED chips,this paper proposed a clustering-based Micro-LED chip micro-image stitching method.Firstly,the OTSU method was used to preprocess the images,and then the adaptive area expansion phase correlation method was used to peak the overlapping regions of two adjacent images to obtain the offset of the preliminary alignment.Subsequently,the optimized alignment strategy using image chunking and the DHash algorithm was employed to obtain the optimized offset.Finally,density-based spatial clustering of applications with noise was introduced to automatically screen out and deal with the possible erroneous offsets.The experimental results show that the average time consumption of this paper's algorithm is 64 ms,which meets the demand of real-time stitching,the stitching correct rate reaches 99.4%,the alignment accuracy is controlled within 1 pixel,and the fusion processing can achieve a perfect stitching effect subjectively,which also provides a new solution to the overall misalignment problem caused by the wrong offsets.The algorithm in this paper effectively solves the key link of microscopic image stitching in wafer-level Micro-LED chip inspection and can be promoted and applied to other machine vision automation fields with repetitive features and high precision requirements,with high practical engineering application value.

关 键 词：图像拼接 Micro-LED芯片显微图像 相位相关 差异值哈希 聚类思想 缺陷检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]