平滑处理的聚类PCA焊缝识别研究  

Research on Weld Recognition of Smooth Clustering PCA

作　　者：谢佳伟 李长勇[1] 伊里哈木·阿布都热木[1] XIE Jia-wei;LI Chang-yong;ILHAM Abdureyim(School of Mechanical Engineering,Xinjiang University,Xinjiang Urumqi 830047,China)

机构地区：[1]新疆大学机械工程学院,新疆乌鲁木齐830047

出　　处：《机械设计与制造》2023年第10期227-232,共6页Machinery Design & Manufacture

基　　金：新疆维吾尔自治区自然科学基金联合项目(2017D01C038)。

摘　　要：针对焊后焊缝识别仍存在适应性不强、精度较低的问题,提出了一种基于平滑处理的聚类PCA焊缝识别算法。首先对焊缝图像灰度化预处理,进行Canny检测,对获取的焊缝边缘进行细节点聚类,在此基础上进行主成分分析,对焊缝的细节点计数,并将结果映射;其次为消除映射数据的杂点,进行一维平滑处理,进行数据的阈值化处理并获取峰值的位置,对峰值域进行求导得到焊缝的左、右边界;然后进行反映射得到焊缝的上下边界,最后基于以上方法设计了四组对实验,重点考虑光照强度、分辨率、焊接方式、焊缝数量对本算法的影响。结果表明,这种焊缝识别算法适用于直型焊缝的识别,其精度高于95%。Aiming at the problems of poor adaptability and low accuracy in weld recognition after welding,a algorithm of cluster-ing principal component analysis(PCA)weld recognition based on smoothing processing was proposed.Firstly,the gray prepro-cessing of the weld image was carried out,and then Canny detection was carried out to cluster the obtained weld edge detail points.On this basis,the principal component analysis was carried out to count the detail points of the weld,and the results were mapped.Secondly in order to eliminate the noise of the mapped data,one-dimensional smoothing is performed,the data is thres-holded and the peak position is obtained,and the peak domain is derived to obtain the left and right boundaries of the weld;then reverse mapping is performed to obtain the upper and lower edges of the weld.In the end,four sets of experiments are designed based on the above methods,focusing on the influence of light intensity,resolution,welding method,and number of welds on the algorithm.The result shows that the welds recognition algorithm is suitable for the recognition of straight welds,and its accuracy is higher than 95%.

关 键 词：Canny检测 主成分分析 一维平滑处理 阈值化 焊缝识别 

分 类 号：TH16[机械工程—机械制造及自动化] TG44[金属学及工艺—焊接] TP18[自动化与计算机技术—控制理论与控制工程]