摘要
利用上下文视觉显著性,探讨了格子、星形和点状色织物瑕疵检测的方法。针对色织物织物背景和瑕疵的对比度和分布特点,提出一种融合主成分分析法和上下文视觉显著性的色织物瑕疵检测算法。针对几种不同图案和纹理的色织物以及5种不同类型的瑕疵进行检测验证。试验结果表明:该算法能够较好地抑制不同类型织物花纹背景,疵点平均检测准确率达92.2%。认为:本文所采用的算法在色织物瑕疵的自动检测应用上具有一定可行性。
With context-aware saliency, the methods for defect inspection of yarn-dyed check fabric, star and dotted types of yarn-dyed fabric were discussed. Aimed at the contrast and distribution characteristics of yarn dyed fabric background and defect,a kind of defect inspection algorithm for yarn-dyed fabric was put forward by combining principle component analysis with context-aware saliency. It was tested and verified based on several yarn-dyed fabrics with different patterns ＆ textures and five kinds of different types of defects. The test results show that the algorithm can well restrain the backgrounds of several types of fabric textures. The average detec- tion rate is 92.2%. It is considered that the algorithm used in the article has feasibility in the automatic detection of defects for yarn-dyed fabrics.
出处
《棉纺织技术》
CAS
北大核心
2018年第2期9-13,共5页
Cotton Textile Technology
基金
国家自然科学基金资助项目(61573095)
上海市自然科学基金资助项目(15ZR1401800)
关键词
色织物
疵点检测
像素
阈值
主成分分析法
最大类间方差法
Yarn-dyed Fabric, Defect Inspection, Pixel, Threshold Value, Principal Component Analysis,OTSU Method
