摘要
针对布匹瑕疵检测中需要无瑕疵图像对照、检测时间过长等问题,提出了一种改进的视觉显著性的布匹瑕疵检测方法,以适用具有不同特征(纹理、颜色、亮度、方向)瑕疵的识别。该方法首先将布匹图像分解为超像素图像,然后根据区域对比度得到布匹瑕疵的显著图,最后对显著图进行阈值分割来识别出布匹瑕疵。实验结果表明本文算法对不同光照下和不同角度下的布匹瑕疵识别有着较高的检测率,具有良好的适用性和鲁棒性。
According to the problem of the contrast of flawless images and the time waste on fabric defect detection,this paper proposed an improved visual significance detection method for fabric flaw detection.The strategy proposed can be used to identify defects with different characteristics( texture,color,brightness,direction). Firstly,this method makes the cloth image decomposing into super-pixel image,then according to the regional contrast it can obtain conspicuous figure of cloth blemish. Finally,the saliency map is segmented by threshold to identify the cloth blemish. The experimental results illuminate that the proposed algorithm has a better detection accuracy and it is more applicable and robust when we recognize the cloth flaws under different illumination and angles.
出处
《信息技术》
2017年第9期22-25,30,共5页
Information Technology
基金
国家自然科学基金(71271117)
江苏省六大人才高峰项目(2013-WLW-005)
江苏省自然科学基金资助(BK20150531)
关键词
布匹瑕疵
瑕疵识别
显著图
超像素
fabric defect
defect recognition
saliency map
super pixel