基于机器视觉的焊点检测算法研究  

Based on Machine Vision Solder Joint Detection Algorithm Research

作　　者：刘美菊[1] 李凌燕[1] 郭文博[1] 

机构地区：[1]沈阳建筑大学信息与控制工程学院,沈阳110168

出　　处：《电子器件》2017年第4期1015-1020,共6页Chinese Journal of Electron Devices

基　　金：辽宁省教育厅项目(L2013225);国家自然科学基金项目(61272253);宁省科学技术研究项目(2014231001)

摘　　要：为了提高电路板焊点检测的准确率,提出了改进的K-近邻法。首先,采用工业相机采集图像并选取470个焊点作为训练样本,利用模板匹配法对图像中的焊点进行定位。然后根据特征分布直方图提取焊点的特征并绘制特征分布情况,选择能区分不同类别焊点的特征作为有效特征。最后,建立改进的K-近邻法焊点检测分类器,选取559个焊点作为测试样本对模型进行测试。实验结果表明改进的K-近邻算法检测的准确率96%以上,可以有效地提高检测效率。In order to improve the detection accuracy of circuit board solder joints, an improvement of K-nearest neighbor method was proposed. Firstly, the industrial camera was used to obtain images, and 470 solder joints were selected as the training samples. The template matching method was used to position the solder joints in the images. Then, the features of solder joints were extracted according to the feature distribution histogram.The space distribu- tion of features was portrayed and then selected the festures can distinguish between solder joints of different types. Finally, the improved K-nearest method classifier was setted up and 559 solder joints were selected as the testing samples to test the model. The experimental results show that the improved K-nearest neighbor method detection accuracy rate reach as high as 96% and can effectively improve the efficiency of detection.

关 键 词：机器视觉 焊点检测 特征提取 改进的K-近邻法 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]