基于贝叶斯模型的印刷电路板ECT图像分割与识别方法  

Image segmentation and recognition method for ECT of printed circuit board based on Bayesian model

作　　者：刘莉[1] 陈玉[1] 程军[1,2] 汪步云 许德章[1,2] Liu Li;Chen Yu;Cheng Jun;Wang Buyun;Xu Dezhang(School of Artificial Intelligence,Anhui Polytechnic University,Wuhu 241000,China;AhpuRobot Industrial Technology Research Institute,Wuhu 241007,China)

机构地区：[1]安徽工程大学人工智能学院,芜湖241000 [2]安普机器人产业技术研究院有限公司,芜湖241007

出　　处：《国外电子测量技术》2022年第9期1-8,共8页Foreign Electronic Measurement Technology

基　　金：教育部重点实验室开放课题(GDSC202011);安徽高校自然科学重点项目(KJ2020A0232);安徽工程大学中青年拔尖人才计划项目;安徽工程大学-鸠江区协同创新专项基金重点项目(2021cyxta1);安徽工程大学创新团队、安徽省重点研发计划(202004a05020013);安徽工程大学科技成果转化引导基金项目资助。

摘　　要：提出了基于高频涡流检测(high frequency eddy current testing, HF-ECT)扫描成像的印刷电路板(printed circuit board, PCB)在线检测方法,利用板中导线、焊盘、引脚与基材之间的电导率差异来提取涡流响应信号并成像,实现对导线断线、焊盘和引脚翘曲、脱落等损伤的高分辨率检测。首先根据线圈激励(频率高达2 MHz)下的涡流分布和成像特性分析印刷电路板的扫描图像特点,并研究图像的染色方法,以达到最佳的像素梯度分布特性;接着,提出基于贝叶斯模型的图像阈值分割和损伤区域提取方法,根据像素灰度分布的后验概率推导出图像分割的最佳阈值,通过对分割后二值图像的边界提取并对其中的损伤进行识别定位;最后将提出的方法应用到PCB的损伤检测中,验证不同条件下的损伤识别和定位效果。研究为印刷电路板的结构表征和损伤检测提供了一种新的途径。An online inspection method for printed circuit board(PCB), based on high frequency eddy current testing(HF-ECT) scan imaging, is proposed in this paper, in which the conductivity difference between the wires, pads, lead foots and the substrate is used to extract the response signals, so as to achieve high resolution detection of defects such as wire breakage, pad warpage and shedding. Firstly, the characteristics of the scanned image of PCB are analyzed, according to the eddy current distribution and imaging characteristics under the high frequency coil excitation with the frequency( up to 2 MHz), and the dyeing method of the image is also studied to acquire the best pixel gradient distribution. Then, the threshold segmentation and defect region extraction method based on Bayesian model is presented. The optimal threshold value for image segmentation is derived from the posterior probability of pixel gray distribution, and the defect positions are identified by extracting the boundary of the binary image after segmentation. Finally, the proposed method is applied to PCB detection, to verify the effect of defect recognition and location under different conditions. This study provides a new approach for structural characterization and damage detection of printed circuit boards.

关 键 词：印刷电路板 涡流检测 贝叶斯模型 阈值优化 损伤识别 

分 类 号：TP274.5[自动化与计算机技术—检测技术与自动化装置]