摘要
为了对具有纹理图案的纺织品缺陷进行准确有效的检测,提出了一种新的基于规则带(regular band,RB)特征的纹理图案布匹瑕疵检测方法。首先利用规则带分析提取图像的各子带纹理特征;然后对各子带图像特征求梯度,并通过融合各子带梯度来获得纹理梯度,使其在纹理梯度中能有效地突出纹理区域的边界;最后在此基础上,结合分水岭分割,准确地检测出纺织品的缺陷区域。实验结果表明,新算法能够准确地定位图案纹理中的瑕疵区域,且操作简便,鲁棒性强。
A novel method for the Patterned Texture Inspection based on regular band is proposed.The properties of texture defects are characterized using the regularity analysis,then the texture gradient from every sub-band is calculated.Texture gradient contains many texture features and the boundary of texture regions in patterned images can be enhanced.Combined with the watershed segmentation,the defective regions in patterned images can be detected accurately.Experimental results show that the method can detect defective regions accurately,and it is easy to operate and robust.
出处
《光学技术》
CAS
CSCD
北大核心
2011年第6期713-717,共5页
Optical Technique
基金
国家自然科学基金资助项目(60804013)
"嵌入式机器视觉关键技术的研究与开发"产学研合作项目资助(011002)
关键词
纹理瑕疵
规则带
纹理分析
分水岭分割
texture defect
regular band
texture analysis
watershed segmentation
