基于AdaBoost和分类树的贴片元件缺陷检测算法  

Defect Detection Algorithm for Patch Components Based on AdaBoost and Classification Tree

作　　者：陈韬 陆艺 李静伟 CHEN Tao;LU Yi;LI Jingwei(College of Metrology and Measurement Engineering,China Jiliang University,Hangzhou 310018,China;Hangzhou Wolei Intelligent Technology Co.,Ltd.,Hangzhou 310018,China)

机构地区：[1]中国计量大学计量测试工程学院,杭州310018 [2]杭州沃镭智能科技股份有限公司,杭州310018

出　　处：《组合机床与自动化加工技术》2024年第10期95-99,104,共6页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：浙江省“尖兵”计划项目(2023C01061)。

摘　　要：针对PCB上贴片元件缺陷检测准确率低、效率低和缺陷类型不全面的问题,设计了一种基于AdaBoost和分类树的贴片元件缺陷检测系统。该系统可检测芯片引脚和电阻缺陷。首先,对采集到的图像进行拼接、校正、元件定位和去噪操作;其次,对贴片元件进行区域划分并提取子区域的形状特征、灰度特征和纹理特征;然后,利用AdaBoost算法将每个特征视为弱分类器,选取最优特征迭代形成强分类器并通过信号函数进行输出,实现每个缺陷都有其对应的特征码;最后,通过查询分类树实现缺陷分类。实验结果表明,相比于传统的图像处理缺陷检测系统,所设计的系统在检测缺陷多样化、检测速度和准确率上均具有明显优势。A PCB surface mount component defect detection system based on AdaBoost and decision trees was designed to address the problems of low accuracy,low efficiency,and incomplete defect types in traditional detection systems.The system detects chip pin and resistor defects.The system first performs image stitching,correction,component positioning,and noise reduction on the collected images.Then,it divides the surface mount components into regions and extracts shape,grayscale,and texture features from each sub-region.The AdaBoost algorithm is used to treat each feature as a weak classifier,select the optimal features to form a strong classifier iteratively,and output a feature code for each defect through a signal function.Finally,the defect classification is achieved by querying the decision tree.Experimental results show that compared with traditional image processing defect detection systems,the system designed in this paper has significant advantages in detecting diverse defects,detecting speed,and accuracy.

关 键 词：机器视觉 印刷电板 图像处理 ADABOOST 分类树 缺陷分类 

分 类 号：TH165[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]