摘要
针对传统裂缝图像信息提取方法的局限性,提出了一种基于多种连通域特征的工程结构表面裂缝提取方法.在采用最大类间方差(Otsu)法对原始图像进行初始分割的基础上,对裂缝图像的连通域面积、最小外接矩形长宽比和连通域内最远距离等连通域特征参数分别进行K-means聚类分析,得到了裂缝目标和噪声背景区域的连通域特征参数分布范围;进而建立了一种新的裂缝信息提取方法,并给出了具体算法流程.该方法与其他方法对比验证表明:该算法计算得到有效性评价系数平均值为0.039 3,比其他方法具有更强的抗噪性和适用性.
According to the situation that traditional methods of extracting crack image information have limitations,a method for extracting surface cracks of engineering structures based on multiple connected domain features was proposed.By using maximum inter-class variance(Otsu)method to split original image initially,K-means clustering algorithm was used to analyze the characteristic parameters of connected domain,including the area of connected domain,minimum length-width ratio of connected domain bounding box and the furthest distance in connected domain in crack image,and the distribution range of characteristic parameters in the connected region of crack target and noise background region were obtained.Then,a new method for extracting crack information was established,and the flow chart of the algorithm was given.The new method was verified by comparing with other methods,and the average validity evaluation coefficient calculated by this algorithm is 0.039 3,which is proved to have better noise resistance and applicability compared with other methods.
作者
徐港
赵恬悦
蒋赏
高德军
XU Gang;ZHAO Tianyue;JIANG Shang;GAO Dejun(School of Civil Engineering and Architecture,China Three Gorges University,Yichang 443002,Hubei China;Hubei Key Laboratory of Disaster Prevention and Mitigation,China Three Gorges University,Yichang 443002,Hubei China)
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2019年第10期52-55,68,共5页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(51379111)
国家重点研发计划资助项目(2017YFC501104)
关键词
混凝土
裂缝
结构检测
聚类分析
图像处理
concrete
crack
structure inspection
clustering analysis
image processing
