一种应用于PCB缺陷检测的改进SIFT算法  

An Improved SIFT Algorithm for PCB Defect Detection

作　　者：黄忠天 陈伟[1] 李昭慧 解文龙 HUANG Zhongtian;CHEN Wei;LI Zhaohui;XIE Wenlong(College of Communication and Information Engineering,Xi an University of Science and Technology,Xi an 710054,China)

机构地区：[1]西安科技大学通信与信息工程学院,陕西西安710054

出　　处：《无线电工程》2023年第6期1479-1486,共8页Radio Engineering

基　　金：国家自然科学基金(61705178)。

摘　　要：针对自动化光学检测设备对印刷电路板(Printed Circuit Boards,PCB)裸板进行缺陷检测的过程中,采集的图像无法兼顾高分辨率与大视角的问题,提出了一种应用于PCB裸板缺陷检测的改进SIFT拼接算法。使用白平衡算法优化图像的质量,提高特征点的检测数量以及改善特征点的分布情况;设计分块策略并利用相似度度量准则的手段,缩小SIFT算法的有效范围,提升算法的整体运行效率;采用RANSAC算法与融合算法完成拼接。实验结果表明,与传统的SIFT+RANSAC算法相比,该算法在拼接准确度上提高了11.3%,拼接效率整体提高了52%,且最终拼接图像的质量也较好。During the process of defect detection of bare PCB boards by automated optical inspection equipment,the captured image cannot combine high resolution and wide viewing angle in the same time.An improved SIFT stitching algorithm for defect detection of bare PCB boards is proposed.Firstly,the white balance algorithm is used to optimize the image quality,increase the number of feature points detected and improve the distribution of feature points.Then the segmentation strategy is designed and the measure of similarity is used to reduce the effective range of SIFT algorithm,improving the overall efficiency of the algorithm.Finally,RANSAC algorithm and fusion algorithm are used to complete the stitching.The experimental results show that compared to the traditional SIFT+RANSAC algorithm,this algorithm improves stitching accuracy by 11.3%and the overall stitching efficiency by 52%.Furthermore,the quality of the final stitched image is also good.

关 键 词：PCB缺陷检测 SIFT算法 白平衡算法 分块策略 相似度度量 图像拼接 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]