摘要
为提高路面裂缝图像分割的准确性与鲁棒性,以学生t分布混合模型(Student\s t-distribution Mixture Model,TMM)代替传统高斯混合模型(Gaussian Mixture Model,GMM),提出一种基于学生t分布混合模型的路面裂缝图像分割方法。所提出方法首先利用K-Means算法来进行初次图像分割。为克服K-Means算法的缺点,以烟花算法(Fireworks Algorithm,FA)运行结果作为K-Means算法初始聚类中心。然后,运行K-Means算法进行初次分割,以此结果作为TMM参数求解的初始值,执行最大期望算法(Expectation-Maximization algorithm,EM),通过交替运行E步和M步,获得模型最终参数值,借助于贝叶斯公式,完成最终图像分割。最后,通过仿真图像和实际图像验证,结果表明该方法分割精度更高、稳定性更好。
In order to improve the accuracy and robustness of pavement crack image segmentation,the Student’s t-distribution Mixture Model(TMM)is used to replace the traditional Gaussian Mixture Model(GMM),and a method for segmentation of pavement crack image based on the TMM is proposed.The proposed method first uses the K-Means algorithm for initial image segmentation.In order to overcome the shortcomings of the K-Means algorithm,the Fireworks Algorithm(FA)operation results are used as the initial clustering centers of the K-Means algorithm.Then,the initial segmentation is worked out by running the K-Means algorithm,the result serves as the initial value of the TMM parameter solution,and the Expectation-Maximization algorithm(EM)is executed,and the E step and M step are runned alternately to obtain the final parameter value of the model.With the help of Bayesian formula,the final image segmentation is completed.Finally,the experimental results of the synthesized image and the actual image show that this study obtains a segmentation result with higher accuracy and better stability than the comparison method.
作者
卢印举
段明义
李祖照
苏玉
LU Yin-ju;DUAN Ming-yi;LI Zu-zhao;SU Yu(School of Information Engineering,Zhengzhou University of Technology,Zhengzhou 450044,China;Guangxi Transportation Research Institute Co.Ltd.,Nanning 530007,China)
出处
《公路》
北大核心
2021年第4期40-47,共8页
Highway
基金
河南省科技攻关计划项目,项目编号192102210120、202102210369
河南省高等学校重点科研项目,项目编号20B520036。
