摘要
为实现织物起毛起球性能评价的客观性及准确性,提出一种基于视觉显著性的织物起球等级评价新算法。首先,利用局部特征显著机制—高斯金字塔分解实现毛球与织物纹理及光照不均等信息的分离,并采用中央-周边操作获得织物起球疵点显著图,再通过融合全局特征显著机制—谱残差提高织物起球疵点显著度;其次,采用平均结构相似度评价算法对织物起球显著图进行质量评估;最后,采用高斯拟合阈值分割得出织物起球兴趣图,从而提取起球特征参数进行等级评价。实验结果表明:该算法能够增强起球显著度,得到有效的起毛起球特征参数,实现1~5级织物起毛起球等级的准确评定。
In order to achieve the objectivity and accuracy of the evaluation of fabric pilling performance,a new objective evaluation of fabric pilling grade based on visual saliency model is proposed.First of all,the local feature saliency mechanism-Gaussian pyramid decomposition was used to achieve that pilling was separated from the uneven information of fabric texture and illumination.And the center-surround operation was used to obtain the saliency images of fabric pilling defects.Then the global feature saliency mechanism-spectral residual method was used to improve the saliency of fabric pilling defects,and mean structural similarity was used to evaluate the quality of fabric pilling saliency images.Finally,the fabric pilling interest images was obtained by Gaussian fitting threshold segmentation,and the pilling feature parameters were extracted for grade evaluation.The results show that the algorithm can enhance the pilling saliency and obtain effective pilling characteristic parameters,and realize the accurate evaluation of 1~5 fabric pilling grade.
作者
管声启
胡璐萍
常江
倪奕棋
王琪璇
管宇灿
张理博
GUAN Shengqi;HU Luping;CHANG Jiang;NI Yiqi;WANG Qixuan;GUAN Yucan;ZHANG Libo(School of Mechanical and Electrical Engineering,Xi’an Polytechnic University, Xi’an 710048, China;Shaoxing Keqiao West-Tex Textile Industry Innovative Institute,Shaoxing 312030, Zhejiang,China;School of Computer and Network Engineering, Shanxi Datong University, Datong 037009,Shanxi,China)
出处
《西安工程大学学报》
CAS
2021年第4期20-27,共8页
Journal of Xi’an Polytechnic University
基金
陕西省重点研发计划项目(2018GY-020)
绍兴市柯桥区西纺纺织产业创新研究院2019年度产学研协同创新项目(19KQYB13)
大学生创新创业训练计划项目(S202010709083)。
关键词
织物起球
高斯金字塔分解
谱残差
结构相似度
中央-周边操作
阈值分割
fabric pilling
Gaussian pyramid decomposition
spectral residual
structural similarity
center-surround operation
threshold segmentation
