基于距离滤波的厚板多层多道焊特征点识别算法  

Recognition algorithm of feature points of multi-pass welding thick plate based on the distance filtering

作　　者：欧志辉 孙振国[1,2,3] 陈咏华[1] 吴景然[1,2] 

机构地区：[1]清华大学先进成形制造教育部重点实验室,北京100084 [2]清华大学天津高端装备研究院,天津300300 [3]浙江清华长三角研究院,浙江嘉兴314006

出　　处：《焊接技术》2017年第5期8-12,共5页Welding Technology

基　　金：国家自然科学基金资助项目(51475259)

摘　　要：针对多层多道焊标准与非标准打底焊缝及各道填充焊缝在激光辅助视觉下的图像信息,提出了一种基于距离滤波的图像处理算法来快速、准确识别出焊缝特征点。该算法首先对激光条纹图像进行二值化、开运算等预处理;对预处理后的图像按列扫描,利用重心法提取激光条纹中心线,针对中心线出现断点的情况,通过直线拟合来填充断点;然后选取中心线上若干点拟合直线,通过求中心线上点到直线的距离,对中心线上的点进行3次滤波处理,滤除中心线上大量与特征点无关的点;最后快速找到中心线上4个端点并进行距离补偿得到焊缝的4个特征点。试验结果表明,该图像处理算法能快速、准确识别出焊缝特征点,为多层多道焊缝跟踪提供了重要依据。For the image information under the laser-assisted vision of multi-pass welding which included standard or non-standard root weld and various weld,we proposed a novel image processing based on the distance filter in order to recognize weld seam feature points quickly and accurately.Firstly,the algorithm preprocessed the laser stripe image by binarization and opening operation,then column scanning the image,extracting the laser stripe centerline by using the center of gravity method,filling the breakpoints on the centerline by using the straight-line fitting method.Next,the algorithm selected many points on the centerline to fit a straight line,and calculated the distance between the line and the centerline points,then triple-filtering processes of the centerline points to filter out all the irrelevant points on the centerline.Finally,the algorithm finded the four endpoints on the centerline quickly and obtained the four feature points by distance compensation.Experimental results show that the image processing algorithm could identify the feature points of weld quickly and accurately.This algorithm provided an important basis for the multi-pass weld tracking.

关 键 词：距离滤波 激光辅助视觉 多层多道焊 图像处理 焊缝跟踪 

分 类 号：TG409[金属学及工艺—焊接]