Markov Random Field Road Crack Image Segmentation Algorithm Integrating Multi-Scale Features

被引：0
|
作者：
Lu Y. [1 ,2 ]
Ma F. [2 ]
Dai S. [1 ]
Su Y. [2 ]
机构：
[1] School of Optical Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai
[2] College of Information Engineering, Zhengzhou University of Technology, Zhengzhou
来源：
Jisuanji Fuzhu Sheji Yu Tuxingxue Xuebao/Journal of Computer-Aided Design and Computer Graphics | 2022年 / 34卷 / 05期
关键词：
Finite mixture model; Image segmentation; Iterated conditional models algorithm; Markov random field; Scale space;
D O I：
10.3724/SP.J.1089.2022.18973
中图分类号：
学科分类号：
摘要：
In order to improve the accuracy and robustness of the pavement crack segmentation, a Markov random field model for crack segmentation that integrates gray-scale statistical models and multi-scale feature vectors is proposed. First, based on the histogram of the crack image, the Gaussian distribution and the Rayleigh distribution are used to frame the crack image gray-scale statistical model, and the EM algorithm is adopted to optimize the model parameters. Then, the Hessian matrix of the crack is constructed by the convolution of the bidirectional Gaussian kernel function and the crack image, and the multi-scale feature vector of the crack image is extracted by calculating the response value of the crack measurement at different scales to enhance the tree-like geometric structure. Finally, the multi-scale feature vectors are integrated into the Markov random field model of crack segmentation, and based on the minimum energy criterion, a conditional iterative algorithm is applied to solve the maximum label field of the crack. On the public data set CrackTree206 and 200 self-built crack data sets, it is compared with 3 algorithms such as MRF. The experimental results show that the probabilistic rand index reaches 91.78%, the global consistency error index reaches 9.86%, and the evaluation index is better than other algorithms, indicating that the algorithm can effectively improve the accuracy of crack segmentation. © 2022, Beijing China Science Journal Publishing Co. Ltd. All right reserved.
引用
收藏
页码：711 / 721
页数：10
相关论文
共 28 条
  • [1] Cao Jingang, Yang Guotian, Yang Xiyun, Pavement crack detection with deep learning based on attention mechanism, Journal of Computer-Aided Design & Computer Graphics, 32, 8, pp. 1324-1333, (2020)
  • [2] Li Gang, Gao Zhenyang, Zhang Xinchun, Et al., Improved global convolutional network for pavement crack detection, Laser & Optoelectronics Progress, 57, 8, pp. 103-111, (2020)
  • [3] Otsu N., A threshold selection method from gray-level histograms, IEEE Transactions on Systems, Man, and Cybernetics, 9, 1, pp. 62-66, (1979)
  • [4] Zou Q, Zhang Z, Li Q Q, Et al., DeepCrack: learning hierarchical convolutional features for crack detection, IEEE Transactions on Image Processing, 28, 3, pp. 1498-1512, (2019)
  • [5] Yang F, Zhang L, Yu S J, Et al., Feature pyramid and hierarchical boosting network for pavement crack detection, IEEE Transactions on Intelligent Transportation Systems, 21, 4, pp. 1525-1535, (2020)
  • [6] Dinh T H, Ha Q P, La H M., Computer vision-based method for concrete crack detection, Proceedings of the 14th International Conference on Control, Automation, Robotics and Vision, pp. 1-6, (2016)
  • [7] Wang G C, Peter W T, Yuan M D., Automatic internal crack detection from a sequence of infrared images with a triple-threshold Canny edge detector, Measurement Science and Technology, 29, 2, (2018)
  • [8] Qin G F, Li Q T., Pavement image segmentation based on fast FCM clustering with spatial information in internet of things, Multimedia Tools and Applications, 78, 5, pp. 5181-5191, (2019)
  • [9] Xiao Y N, Li J Y., Crack detection algorithm based on the fusion of percolation theory and adaptive canny operator, Proceedings of the 37th Chinese Control Conference, pp. 4295-4299, (2018)
  • [10] Tian B Z, Xu S N, Li Z H., Glassware crack defects detection based on wavelet transform, Proceedings of the 17th Chinese Automation Congress, pp. 4954-4958, (2017)
123