焊缝缺陷三维成像及亚像素定量方法研究

Research on 3D imaging and sub-pixel quantitative method of weld defects

为实现工业设备中焊缝结构内部缺陷准确地定量分析,针对构造缺陷三维重构模型时缺陷图谱的过分割和欠分割导致定量分析不够精确的问题,提出了一种基于图谱颜色分割的新算法。利用该算法处理缺陷切片并进行三维重构获取其长度,应用Canny算子与三次样条插值法结合获取缺陷亚像素边缘,基于缺陷区域连通标记算法和累加法可获取缺陷面积和体积的方法。研究结果表明:与实际数据对比,未焊透、未熔合、夹渣缺陷重构的长度误差分别为8.18%、5%、9.62%,基于此对分割缺陷的亚像素边缘内连通域进行标记以准确地获取其面积及体积。该方法能够在不破坏焊缝结构的情况下即时获取焊缝内部缺陷的形貌并对其进行精准且全面的定量分析。

In order to achieve accurate and quantitative analysis internal defects of welds which are used in industrial equipment,a segmentation algorithm based on image’s color was proposed to process defect slices and obtain their length through rebuilding their three-dimensional model,aiming at the problem of insufficient accuracy in quantitative analysis caused by over segmentation and under segmentation of the defect images when constructing a three-dimensional model for defects.Then,Canny operator with cubic spline interpolation was used to obtain subpixel edges of defects.Finally,a method for obtaining area and volume of defects based on Connected Component Labelling and accumulation method is proposed.The research results show that:compared with actual data,errors in length about lacking of penetration,lack of fusion and slag defects in three-dimensional model are 8.18%,5%,and 9.62%.Regions within the subpixel edge of defects are labeled to accurately obtain their area and volume.This method can instantly show internal defects’shape and position without damaging weld’s structures,which lay a foundation for analyzing their area and volume.