纺织学报 ›› 2014, Vol. 35 ›› Issue (11): 62-0.

• 纺织工程 • 上一篇    下一篇

基于局部统计与整体显著性的织物疵点检测算法

  

  • 收稿日期:2013-10-28 修回日期:2014-05-17 出版日期:2014-11-15 发布日期:2014-11-20
  • 通讯作者: 刘洲峰 E-mail:lzhoufeng@hotmail.com
  • 基金资助:

    国家自然科学基金;河南省基础与前沿技术研究计划

Fabric defect detection algorithm using local statistic features and global saliency analysis

  • Received:2013-10-28 Revised:2014-05-17 Online:2014-11-15 Published:2014-11-20
  • Contact: Zhou-Feng LIU E-mail:lzhoufeng@hotmail.com

摘要: 由于织物疵点类别较多及图像纹理多样化,为了能更有效检测织物疵点,本研究结合局部统计特征与整体显著性分析,提出一种新的织物疵点检测算法。首先将图像分为大小相同的图像块,采用局部二进制模式和灰度直方图分别提取图像块局部统计特征;其次针对每个当前图像块,随机选取K个其它图像块,分别计算局部二进制模式统计特征对比度和灰度统计特征对比度,完成基于上下文整体显著性分析生成视觉显著图;最后采用基于迭代最优阈值分割算法对显著图进行分割,得到织物疵点检测结果。实验结果表明,该算法综合了局部统计特征和整幅图像的上下文信息,可显著突出织物疵点区域,实现对织物疵点的有效检测。

Abstract: In order to efficiently detect defect for fabric image with complex texture and variety of defects, this paper proposed a novel defect detection algorithm based on local statistical features and global saliency analysis. In the proposed algorithm, the target image is first divided into blocks with the same size, then the LBP technique is used to extract the texture features of the blocks and the histogram technique is used to extract the grayscale statistical features of the blocks. Second, for a given image block, K blocks are randomly chosen for calculating the LBP feature contrast and grayscale histogram feature contrast between the given block and the randomly-chosen blocks. Based on the obtained global contrast information, a saliency map is produced. Finally, the saliency map is segmented by using an optimal threshold, which is obtained by an iterative approach. Through these procedures, the detection result is obtained. The experimental results demonstrate that the proposed algorithm, integrating the local textual and grayscale statistical features and the global saliency analysis, can detect the fabric defections effectively.

中图分类号: 

  • TP391.9
[1] 丁淑敏, 李春雷, 刘洲峰. 基于小波变换及投影分析的织物倾斜度检测算法[J]. 纺织学报, 2012, 33(8): 59-65.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!