作者：卢印举1,2, 豆艳艳2, 戴曙光1, 苏玉2
作者单位：1. 上海理工大学光电信息与计算机工程学院，上海 200093;
2. 郑州工程技术学院信息工程学院，河南 郑州 450044
关键词：图像分割;梯度矢量流;混合模型;马尔科夫随机场;条件迭代算法
摘要：
针对裂缝形态多样性和分布随机性使得传统裂缝图像分割算法的泛化能力弱的问题，提出一种基于几何结构测度的路面裂缝图像分割算法。首先，采用瑞利分布和高斯分布对裂缝图像背景和目标进行建模，并用期望最大化算法求解裂缝灰度混合模型参数；然后，通过高斯核函数与图像的卷积计算裂缝的边界映射，用梯度矢量流场构造裂缝图像Hessian矩阵，由Hessian矩阵描述裂缝测度函数并获取裂缝多尺度特征向量；最后，将裂缝多尺度特征向量和灰度混合模型融合到马尔可夫随机场，基于最小能量准则，利用条件迭代算法求解裂缝最大标号场来实现裂缝图像分割。实验表明，与仅依靠灰度特征的传统裂缝图像分割算法相比，所提算法综合指标达88.02%、重叠率达54.92%，优于其他算法，具有良好的噪声抑制能力和泛化能力。
A road crack image segmentation algorithm based on geometric structure measurement
LU Yinju1,2, DOU Yanyan2, DAI Shuguang1, SU Yu2
1. School of Optical Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China;
2. College of Information Engineering, Zhengzhou University of Technology, Zhengzhou 450044, China
Abstract: Aiming at the problems that the diversity of crack morphology and distribution randomness make traditional crack image segmentation algorithms have poor generalization ability, a road crack image segmentation algorithm based on geometric structure measurement was proposed. First, Rayleigh distribution and Gaussian distribution was adopted to model the crack image background and target, and the expectation maximization algorithm was used to solve the crack gray-scale mixed model parameters. Then, the boundary mapping of the crack was calculated by the convolution of the Gaussian kernel function and the image, the Hessian matrix of the crack image was constructed with the gradient vector flow field. Hessian matrix was adopted to describe crack measurement function and obtain crack multi-scale eigenvectors. Finally, the crack multi-scale eigenvectors and gray-scale hybrid model were merged into the Markov random field. Based on the minimum energy criterion, a conditional iterative algorithm was used to solve the maximum crack label field to realize crack image segmentation. Experiments show that compared with the traditional crack image segmentation algorithm that only relies on gray features, the proposed algorithm has a comprehensive index of 88.02% and an overlap rate of 54.92%, which is better than other algorithms and has good noise suppression and generalization capabilities.
Keywords: image segmentation;gradient vector flow;mixture model;Markov random field;ICM algorithm
2022, 48(4):77-84 收稿日期: 2020-12-17;收到修改稿日期: 2021-02-17
基金项目: 河南省科技攻关计划项目(222102210222)；郑州市科技局基础研究及应用基础研究项目(zkz202103，zkz202105)
作者简介: 卢印举（1976-），男，江苏新沂市人，副教授，博士，研究方向为机器视觉检测
参考文献
[1] YANG F, ZHANG L, YU S, et al. Feature pyramid and hierarchical boosting network for pavement crack detection[J]. arXiv preprint arXiv:1901. 06340, 2019, 15(62): 3708-3712
[2] OTSU N. A threshold selection method from gray-level histograms[J]. IEEE Transactions on Systems Man & Cybernetics, 2007, 9(1): 62-66
[3] AKAGIC A, BUZA E, OMANOVIC S, et al. Pavement crack detection using Otsu thresholding for image segmentation[C]//2018 41st International Convention on Information and Communication Technology, Electronics and Microelectronics (MIPRO). IEEE, 2018: 708-716.
[4] WANG Q, ZHANG L, BERTINETTO L, et al. Fast online object tracking and segmentation: a unifying approach[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Long Beach California, USA, 2019: 1328-1338.
[5] QIN G, LI Q. Pavement image segmentation based on fast FCM clustering with spatial information in internet of things[J]. Multimedia Tools and Applications, 2019, 78(5): 5181-5191
[6] XIAO Y N, LI J Y. Crack detection algorithm based on the fusion of percolation theory and adaptive canny operator[C]//37th Chinese Control Conference (CCC), New York: IEEE, 2018: 4295-4299.
[7] 王昕, 赵飞, 蒋佐富, 等. 迁移学习和卷积神经网络电力设备图像识别方法[J]. 中国测试, 2020, 46(5): 108-113
[8] 殷鹰, 谢罗峰, 黄泰博. 基于深度学习的磁瓦内部缺陷声振检测方法[J]. 中国测试, 2020, 46(3): 32-38
[9] FAN Z, WU Y, LU J, et al. Automatic pavement crack detection based on structured prediction with the convolutional neural network[J]. Computer-Aided Civil and Infrastructure Engineering, 2018, 561(254): 128-150
[10] LAU S L H, CHONG E K P, YANG X, et al. Automated pavement crack segmentation using U-Net-based convolutional neural network[J]. IEEE Access, 2020, 8(7): 114892-14899
[11] DEMPSTER A P, LAIRD N M, RUBIN D B. Maximum-likelihood estimation from incomplete data via the EM algorithm(with discussion)[J]. Journal of the Royal Statistical Society. Series B:Methodological, 1977, 39(1): 1-38
[12] XU C, PRINCE J. Snakes, shapes, and gradient vector flow[J]. IEEE Transactions on Image Processing, 1998, 7(3): 359-369
[13] LIN M, ZHOU R, YAN Q, et al. Automatic pavement crack detection using HMRF-EM algorithm[C]//2019 International Conference on Computer, Information and Telecommunication Systems (CITS), 2019, 44(2): 502-513.
[14] 李鹏, 李强, 马味敏, 等. 基于K-means聚类的路面裂缝分割算法[J]. 计算机工程与设计, 2020, 41(11): 3143-3147
[15] 王森, 伍星, 张印辉, 等. 基于深度学习的全卷积网络图像裂纹检测[J]. 计算机辅助设计与图形学学报, 2018, 30(5): 859-867
