基于视觉显著性的平纹织物疵点检测  

Defect detection of plain weave based on visual saliency mechanism

作　　者：管声启[1] 高照元 吴宁[1] 徐帅华 

机构地区：[1]西安工程大学机电工程学院,陕西西安710048

出　　处：《纺织学报》2014年第4期56-61,共6页Journal of Textile Research

基　　金：陕西省教育厅科研计划项目(2013JK1083);西安工程大学博士科研启动基金项目(BS1005)

摘　　要：受检测环境及疵点特点影响,传统的检测算法难以满足疵点动态检测,为此,提出基于视觉显著性疵点动态检测的新方法。先将采集的图像进行特征提取形成特征图,再对特征图进行小波多层分解形成特征子图;在此基础上,对分解后的特征子图进行中央周边操作构建特征差分子图;然后,通过特征差分子图的融合策略形成显著图;最后,采用阀值法分割出兴趣区,并通过区域生长分割出疵点目标。结果表明,该方法能够完整检测出平纹织物疵点信息,并且具有较强的抗干扰能力。Because of the influence of inspection environment and defect characteristics, the conventional detection methods are difficult to meet the requirements of dynamic detection of defects. Therefore, a new algorithm is designed for dynamic detection based on visual saliency mechanism. First of all, the acquired image features are extracted, thus feature maps are obtained. Secondly, the characteristic sub-maps are formed by wavelet multilevel decomposition of feature maps. On this basis, the center-surround difference operation is applied to the construct characteristic difference sub-maps. Then, the fusion strategy for feature difference sub-maps is used to form salient maps. Finally, the detection interest region is formed by using threshold method, and defect targets are segmented by the region growing. Experimental results show that this method can detect plain weave fabric defect information, and has strong anti-jamming ability.

关 键 词：视觉显著性机制 特征提取 显著图 织物疵点检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]