摘要
为了快速准确地获取不规则曲线对象的中心线长度,提出一种基于机器视觉的测量方法。首先提取对象中心线:对图像二值化后,填充孔洞、去除背景噪点及边沿毛刺以排除干扰,结合Zhang-Suen细化算法和Hilditch细化算法提取中心线骨架,并逐一处理分叉点以去除骨架毛刺,得到单像素宽度且无毛刺的中心线,过程中通过构建特别处理以显著提高算法效率。然后测量中心线长度:利用数字曲线反向多边形逼近算法获取中心线的支撑点序列并进行插值扩充,采用B样条曲线分段拟合支撑点序列。最终通过数值计算方法得出测量结果。使用该方法对分别在15种不同弯曲状态下的5类长度为100 mm的曲线对象进行测试,测量的平均相对误差为0.297%,平均运行时间为0.673 s,相对现有方法在精度和效率两方面均有显著提升,具有一定的应用价值。
To obtain the centerline length of irregular curve objects quickly and accurately,a measurement method based on machine vision is proposed.Firstly,extract the centerline of the object:fill holes,remove background noise and edge burrs in the binarized image to eliminate interference;combine Zhang-Suen and Hilditch refinement algorithm to extract the centerline skeleton;process the fork points one by one to remove skeleton burrs,and obtain a single pixel wide and burr free centerline.During this period,special processing is constructed to improve algorithm efficiency.Secondly,measure the length of the centerline:use inverse polygon approximation of digital curves algorithm to obtain the sequence of support points on the centerline and perform interpolation expansion;segmentally fit the sequence of support points with B-spline curve,and finally obtain the measurement results through numerical calculation method.Using 5 types of curve objects with a length of 100 mm under 15 different bending states to test this method,the average relative error is 0.297%,and the average running time is 0.673 s.This method has significant improvement in both accuracy and efficiency compared with existing methods and has certain application value.
作者
易焕银
彭志良
马旭东
YI Huanyin;PENG Zhiliang;MA Xudong(Guangdong Communication Polytechnic,Guangzhou 510650,Guangdong,China)
出处
《广东交通职业技术学院学报》
2025年第2期64-70,共7页
Journal of Guangdong Communication Polytechnic
基金
广州市基础与应用基础研究项目(编号:202201011675)
广东交通职业技术学院校级教科研项目(编号:GDCP-ZX-2024-024-N1)
广东交通职业技术学院校级大学生科技创新项目(编号:GDCP-ZX-2024-020-N2)。
关键词
机器视觉
不规则曲线
中心线提取
B样条拟合
长度测量
machine vision
irregular curve
centerline extraction
B-spline fitting
length measurement
