汽车车身焊点质量智能检测系统设计  

Research on Detection of Automobile Components Solder Joint Quality

作　　者：林川[1,2] 吕晓峰[3] 曹以隽 张军利[3] 

机构地区：[1]广西科技大学电气与信息工程学院,广西柳州545006 [2]广西科技大学广西汽车零部件与整车技术重点实验室,广西柳州545006 [3]宝鸡文理学院电子电气工程系,陕西宝鸡721013

出　　处：《计算机测量与控制》2014年第1期31-33,共3页Computer Measurement &Control

基　　金：广西汽车零部件与整车技术重点实验室(广西科技大学)开放基金资助(2012KFMS09);广西重点实验室建设项目(13-051-38)

摘　　要：为提高汽车车身焊点质量的自动检测效率,研究基于机器视觉技术的自动检测系统;设计了以TMS320DM648DSP为处理器的系统硬件框图,针对传统的以局部灰度特性作为分类特征检测效果不佳的问题,提出采用具有生物视觉特性的Gabor滤波器对待测图像进行多方向和多尺度滤波,将滤波结果进行特征融合后降维的特征提取方法,将文章提取的各不同维度特征与传统方法提取的特征,分别采用支持向量机(SVM)与BP神经网络、AdaBoost分类器进行检测对比研究;实验结果表明,文章采用的SVM分类器平均检测率达97.73%,具有较好的鲁棒性。In order to enhance the solder joint quality detection rate of automobile body, the auto detection system based on machine vi sion technology was researched. The diagrams of system hardware which used TMS320DM648 DSP as processor were designed. Because the detection results are bad by using local gray characters as the classification features, the Gabor filters which have biological visual characteristic was used to filter the detected images with multiple directions and scales, and then the filter results with feature fusion was processed. The different dimensions features extracted by this paper and the features extracted by traditional methods were all detected and compared by Support Vector Machine （SVM）, Back Propagation net and AdaBoost classifiers. The average detection rate which uses SVM as classifiers is 97. 73% and the method has better robustness.

关 键 词：汽车车身 焊点质量检测 DSP GABOR滤波器 分类器 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]