船体零件几何尺寸测量图像分割方法研究  

Research on Image Segmentation for Geometric Dimension Measurement of Hull Parts

作　　者：王建新[1] 朱煜[1,2] 胡小锋 张亚辉[1] WANG Jianxin;ZHU Yu;HU Xiaofeng;ZHANG Yahui(School of Mechanical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China;Jiangnan Shipyard(Group)Co.,Ltd.,Shanghai 201913,China)

机构地区：[1]上海交通大学机械与动力工程学院,上海200240 [2]江南造船(集团)有限责任公司,上海201913

出　　处：《机械与电子》2021年第5期68-73,共6页Machinery & Electronics

基　　金：国家重点研发计划资助项目(2018YFB1700500)。

摘　　要：针对船体零件表面划痕、锈蚀以及下表面成像等因素导致的伪边缘及边缘不连续问题,提出了基于边缘与k-means聚类的船体零件图像分割方法。对于表面划痕及锈蚀形成的伪边缘,在矩形边缘检测区域内利用轮廓法线方向与边缘梯度方向信息予以剔除;零件下表面成像导致的伪边缘,选取k-means聚类区分零件的上下表面边缘;边缘不连续问题,基于测地距离和断点方向准确连接断点,形成封闭轮廓。实验结果表明,使用该方法可获得平滑、封闭的分割效果,船体零件的几何尺寸测量误差低于0.5 mm,且测得量值标准偏差的平均值为0.262 mm,测量重复性较高,有效保障了船体零件几何尺寸的测量精度。Focused on the issue of pseudo edge and edge discontinuity caused by scratches,corrosion and imaging of undersurface of hull parts,an image segmentation method based on edge detection and edge linking is proposed.The normal direction and edge gradient direction information of the contour are used to eliminate the pseudo edges formed by scratches and corrosion in the rectangular edge detection area;for pseudo edges caused by imaging of the lower surface of parts,the k-means clustering is selected to distinguish the upper and lower surface edges;aiming at the problem of edge discontinuity,based on the geodesic distance and the direction of the break point,the break points are accurately connected to form a closed contour.Experimental results show that the proposed method can obtain smooth and closed segmentation results,the measurement error of hull parts is less than 0.5 mm,and the average value of the measured standard deviation is 0.262 mm.This method effectively ensures the measurement precision of hull parts.

关 键 词：图像分割 边缘检测 K-MEANS聚类 轮廓提取 边缘连接 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]