摘要
针对织物疵点检测过程中疵点种类繁多、大小差异和受织物本身纹理干扰等特点,提出了一种全局显著性和局部显著性相结合的织物疵点检测方法.首先对图像进行稀疏表示,然后计算系数矩阵的增量编码长度,根据增量编码长度量得到局部显著图,再利用频率调谐法计算全局显著图,接着将这两类显著图相融合得到综合显著图,最后,通过自适应阈值分割法求得二值化图像.实验证明:该算法的检测效率高,并具有较强的抗干扰能力.
Aiming at the problems, such as a wide range of defects, different defect sizes and fabric texture interferences in the process of fabric defect detection, this paper proposes a new fabric defect detection method combining the local saliency and global saliency. Firstly, the image is represented as sparse matrix and the incremental encoding length of the coefficient matrix is computed. The local saliency can be measured by the Incremental Coding Length（ICL）. Then the frequency tuning method is used to calculate the global saliency map. These two kinds of saliency maps are combined to obtain the final saliency map. Finally, the adaptive threshold segmentation method is used to achieve binary image. Experiment shows that the proposed algorithm can efficiently detect defects with shorter time and has strong anti-jamming ability.
作者
姚明海
潘海飞
王宪保
YAO Minghai PAN Haifei WANG Xianbao(College of Information Engineering, Zhejiang University of Technology, Hangzhou 310023, China)
出处
《浙江工业大学学报》
CAS
北大核心
2017年第1期19-22,共4页
Journal of Zhejiang University of Technology
关键词
疵点检测
频率调谐
显著性
稀疏编码
defect detection
frequency tuned
saliency detection
sparse coding
