基于纹理结构异常的织物疵点检测算法研究  

A Novel Fabric Defect Detection Method Based on Texture Structure

作　　者：刘洲峰[1] 赵全军[1] 李春雷[1] 董燕[1] 闫磊[1] 

机构地区：[1]中原工学院,郑州450007

出　　处：《中原工学院学报》2014年第3期22-25,共4页Journal of Zhongyuan University of Technology

基　　金：国家自然科学基金项目(61379113);河南省基础与前沿技术研究计划项目(132300410163)

摘　　要：提出了一种基于织物纹理结构异常的织物疵点检测算法。首先计算代表正常纹理的主邻域结构图;其次通过比较每个像素的邻域结构图与主邻域结构图的差异来定义该像素的显著性,进而得到显著图;最后采用迭代最优阈值分割方法对显著图进行分割得到疵点区域。试验证明,该算法操作简单、计算速度快、鲁棒性强,具有较强的自适应性。Based on the structure of the fabric texture characteristics, this paper proposes a fabric defect detection algorithm based on global fabric texture feature. First, the global representative neighborhood struc- ture map is calculated- second each pixel neighborhood structure map are compared with the obtained global neighborhood structure map to define the saliency of the pixel, and then a saliency map is got in the end, the saliency map is segmented by iterative threshold segmentation algorithm to get the defect area. Experiment re- sults demonstrate the effectiveness of the proposed method.

关 键 词：织物疵点 疵点检测 纹理结构 阈值分割 

分 类 号：TM615[电气工程—电力系统及自动化]