Computer Engineering and Applications ›› 2018, Vol. 54 ›› Issue (19): 172-178.DOI: 10.3778/j.issn.1002-8331.1706-0166

Previous Articles     Next Articles

Fuzzy C-Means clustering algorithm under noise uneven condition and its application

WANG Wenhui1, YANG Geng1, GE Wei1, LIU Peidong2, QIAN Chen3   

  1. 1.Jiangsu Key Laboratory of Big Data Security & Intelligent Processing, School of Computer Science, Nangjing University of Posts and Telecommunications, Nanjing 210000, China
    2.Jiangsu Hengtong Au Optronics Co., Suzhou, Jiangsu 215200, China
    3.School of Optoelectronic Engineering, Nangjing University of Posts and Telecommunications, Nanjing 210000, China
  • Online:2018-10-01 Published:2018-10-19


王文慧1,杨  庚1,葛  炜1,刘沛东2,钱  晨3   

  1. 1.南京邮电大学 计算机学院 江苏省大数据安全与智能处理重点实验室,南京 210000
    2.江苏亨通光电有限公司,江苏 苏州 215200
    3.南京邮电大学 光电学院,南京 210000

Abstract: With the industrial production and technological progress, people on the product quality requirements are getting higher and higher. In order to improve the effect of Fuzzy C-Means clustering algorithm on noise-sensitive fragmentation, a new Fuzzy C-Means(FCM) clustering defective image segmentation method is proposed. The method considers the neighborhood pixel information of the sample on the one hand, the FCM membership function contains the neighborhood information of the pixel, and the other uses a new distance metric instead of the traditional Euclidean distance. The above two methods are used to increase the robustness of the algorithm. In addition, through the histogram method to the cluster center by the initial value, so that the segmentation effect is stable. Finally, the pretreatment of the cable image is added by using the improved FCM algorithm and the traditional FCM algorithm and FCMM algorithm to add the salt and pepper noise and the Gaussian white noise respectively to the cable image acquired by the CCD camera. The comparison of the experimental results and the correctness of the segmentation show that the improved FCM algorithm can better overcome the noise and extract the flaws from the image accurately. The defect profile is clearer and the effect of the flaw detection on the surface of the cable is improved.

Key words: fiber-optic cable surface, segmentation of defects, Fuzzy C-Means clustering, sample neighborhood pixel, Euclidean distance

摘要: 随着工业生产和工艺的进步,人们对产品的质量要求越来越高。为提高光缆表面瑕疵分割的效果,克服模糊C均值聚类算法对噪声敏感的不足,提出了一种新的模糊C均值聚类(FCM)的瑕疵图像分割方法。该方法一方面考虑样本的邻域像素信息,使FCM的隶属度函数中包含像素的邻域信息,另一个方面使用一种新的距离度量方式代替传统的欧式距离。利用以上两种方法来增加算法的鲁棒性,此外,通过直方图法给聚类中心赋初值,使分割效果稳定。最后,分别对CCD相机获取的光缆图像添加椒盐噪声和高斯白噪声,使用改进的FCM算法和传统的FCM算法、FCMM算法进行光缆表面瑕疵分割实验。图像和分割正确率的对比实验结果表明,使用改进的FCM算法能更好地克服噪声,精确地将瑕疵从图像上提取出来,瑕疵轮廓更为清晰,提高了光缆表面瑕疵检测的效果。

关键词: 光缆表面, 瑕疵分割, 模糊C均值聚类, 样本邻域像素, 欧式距离