电焊机2025，Vol.55Issue(5)：100-106,7.DOI:10.7512/j.issn.1001-2303.2025.05.14
高层建筑钢结构梁柱节点焊接裂纹缺陷图像检测改进
Improvement of Image Detection for Welding Crack Defects in Steel Structure Beam and Column Nodes of High rise Buildings
摘要
Abstract
To address the issue of approximate interference of boundary pixels in UAV-captured images of welding crack de-fects at steel structure beam-column joints in high-rise buildings,this study proposes an improved detection method based on the concept of deriving approximate solutions for pixel boundary values of similar crack defects.Firstly,the grayscale mean of the original image is calculated,and the mean parameters of different regions are adjusted to enhance the average grayscale.Subsequently,a J-integral linear formula is introduced for suspected crack defect areas in steel beam-column nodes,from which a pixel gradient formula is derived to identify the general defect regions.The linear relationship between defect pixel regions and crack intensity factors is established to determine boundary gradient regions of defect points.Fi-nally,by setting the maximum pixel gradient of normal points and calculating pixel variances in high-gradient and low-gradient regions,welding crack defects are detected through variance comparisons.Experimental results demonstrate that the proposed method achieves a detection accuracy consistently above 95%,peaking at 98%,representing an approximately 20%improvement over traditional convolutional neural network methods.This research provides a high-precision,low-complexity solution for non-destructive testing of steel structure welding quality,with significant potential for engineering applications.关键词
钢结构梁柱节点/焊接裂纹缺陷/J积分线性公式/像素最值梯度/低梯度区域Key words
steel structure beam column nodes/welding crack defects/J-integral linear formula/pixel maximum gradient/low gradient region分类
金属材料引用本文复制引用
王力军..高层建筑钢结构梁柱节点焊接裂纹缺陷图像检测改进[J].电焊机,2025,55(5):100-106,7.基金项目
山东省技术创新项目(202350101362) （202350101362）
