基于Canny算子边缘分割的海洋平台焊接构件焊缝检测方法  

Welding seam detection method for ocean platform welding components based on Canny operator edge segmentation

作　　者：顾天宝 GU Tianbao(Offshore Oil Engineering Co.,Ltd.,Tianjin 300452,China)

机构地区：[1]海洋石油工程股份有限公司,天津300452

出　　处：《无损检测》2024年第10期70-75,共6页Nondestructive Testing

摘　　要：当前海洋平台焊接构件焊缝的检测过程中,依托于Sobel(索贝尔)算子完成图像边缘分割,对噪声信息比较敏感,使得检测结果ACC值(准确率)较低。因此,提出基于Canny算子边缘分割的海洋平台焊接构件焊缝检测方法。运用数学形态学中的膨胀和腐蚀思想,对海洋平台焊接构件焊缝图像进行滤波处理。首先将预处理后的图像输入视觉注意模型,提取图像包含的综合显著性特征,从而标注出图像视觉显著区域;然后基于Canny算子边缘分割算法,检测出图像中的全部边缘像素,并通过膨胀和细化操作实现边缘连接,从图像中分割出整个目标区域;最后选定一个种子区域,按照生长判定准则获取区域生长结果,从而得出焊缝检测结果。试验结果表明,新设计方法的焊缝检测结果ACC值稳定在0.95以上,能够满足海洋平台焊接构件的安全检测要求。In the current welding seam detection process of offshore platforms,image edge segmentation is achieved by relying on the Sobel operator,which is sensitive to noise information,resulting in lower ACC values(accuracy)of the welding seam detection results.Therefore,a welding seam detection method for offshore platform welded components based on Canny operator edge segmentation was proposed.Using the concepts of expansion and corrosion in mathematical morphology,the welding seam images of offshore platform welding components were filtered.The preprocessed images were input into the visual attention model,the comprehensive saliency features contained in the image were extracted,and the visual salient regions of the image were annotated.Based on the Canny operator edge segmentation algorithm,all edge pixels in the image were detected,and edge connections were achieved through dilation and refinement operations to segment the entire target area from the image.A seed region was selected and the region growth results were obtained according to the growth judgment criteria,in order to obtain the weld seam detection results.The experimental results showed that the ACC value of the new design method for weld seam detection was set above 0.95,which well met the safety inspection requirements of welded components on offshore platforms.

关 键 词：CANNY算子 边缘分割 图像处理 海洋平台 区域生长 焊缝检测 

分 类 号：TP21[自动化与计算机技术—检测技术与自动化装置] TG115.28[自动化与计算机技术—控制科学与工程]