考虑缺陷随机特征的PCB裸板缺陷图像检测  

PCB Bare Board Defect Image Detection Considering Random Characteristics of Defects

作　　者：张婵 ZHANG Chan(School of Information Technology,Guangdong Industry Polytechnic University,Guangzhou 510300,China)

机构地区：[1]广东轻工职业技术大学信息技术学院,广州510300

出　　处：《微处理机》2024年第5期42-45,共4页Microprocessors

基　　金：2023年中国高校产学研创新基金(编号:2022 IT233)。

摘　　要：为提高PCB裸板缺陷检测精度,提出一种考虑缺陷随机特征的新方法。该方法对PCB裸板图像进行预处理,包括灰度化、去噪、划分边界和轮廓;通过定点处理分离缺陷区域和非缺陷区域,提取缺陷随机特征;利用支持向量机进行缺陷分类,并使用随机森林回归模型定位缺陷。针对随机选定的多个PCB板元件位置进行测试,结果显示该方法能有效检测PCB裸板表面缺陷,在多个随机选定的PCB板元件位置上具有较高的检测准确率和较低的漏检率,各阶段检测平均精度均较高,可在PCB生产质量控制过程中发挥重要的实际作用。A new method considering random characteristics of defects is proposed to improve PCB bare board defect detection accuracy.The method preprocesses PCB bare board images,including grayscale conversion,denoising,boundary and contour division.It separates defect and non-defect areas through fixed-point processing,extracting random defect features.Support Vector Machine is used for defect classification,while Random Forest regression model locates defects.Tests on randomly selected PCB component positions show that the method effectively detects PCB bare board surface defects,achieving high detection accuracy and low miss rate across multiple randomly selected PCB component positions.With high average detection precision at various stages,it can play a significant practical role in PCB production quality control.

关 键 词：PCB裸板 中值滤波 CANNY边缘检测 支持向量机 随机森林回归 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]