混合色彩空间与多核学习的色纺织物组织点识别  

Recognition of colored spun fabric interlacing point based on mixed color space and multiple kernel learning

作　　者：龚雪 袁理 刘军平[3] 杨亚莉 刘沐黎 柯政涛 鄢煜尘 GONG Xue;YUAN Li;LIU Junping;YANG Yali;LIU Muli;KE Zhengtao;YAN Yuchen(School of Electronic and Electrical Engineering,Wuhan Textile University,Wuhan,Hubei 430200,China;State Key Laboratory for Hubei New Textile Materials and Advanced Processing Technology,Wuhan Textile University,Wuhan,Hubei 430200,China;School of Mathematics and Computer Science,Wuhan Textile University,Wuhan,Hubei 430200,China;Electronic Information School,Wuhan University,Wuhan,Hubei 430072,China)

机构地区：[1]武汉纺织大学电子与电气工程学院,湖北武汉430200 [2]武汉纺织大学湖北省纺织新材料与先进加工技术省部共建国家重点实验室培育基地,湖北武汉430200 [3]武汉纺织大学数学与计算机学院,湖北武汉430200 [4]武汉大学电子信息学院,湖北武汉430072

出　　处：《纺织学报》2020年第5期58-65,共8页Journal of Textile Research

基　　金：湖北省自然科学基金项目(2014CFB754);湖北省教育厅科学技术研究计划青年人才项目(Q20141607);中国纺织工业联合会科技项目(2018035,2014072)。

摘　　要：针对色纺织物组织点参数特征提取困难的问题,建立了基于混合色彩空间与多核学习的色纺织物组织点自动识别算法。首先,将YUV、HSV和Lab 3种色彩空间中具有相同颜色属性的分量通道进行独立融合,并构建混合色彩空间;在此基础上,分别提取色纺织物组织点图像的局部纹理统计特征与三阶颜色矩特征,用于织物组织点特征参数的表征;最后,通过多核学习算法构建支持向量机,实现织物组织点特征的识别。实验结果表明,所建立的色纺织物组织点识别算法,不仅能够对府绸、斜纹与缎纹等典型结构的组织点进行有效识别,而且对于纤维种类、成纱工艺与织物组分的调整也具有理想的鲁棒性与普适性,其平均识别率达到91. 2%。Aiming at the difficulty in extracting feature parameters of colored fabric interlacing points,an automatic recognition algorithm for such interlacing points based on mixed color space and multiple kernel learning was established. Firstly,the channel having the same color properties among the three-color spaces of YUV,HSV and Lab was fused to construct a mixed color space. On this basis,the local texture features and the third-order color moment features of the image of colored fabric interlacing points were extracted to represent the interlacing points. Finally,support vector machine was constructed by multi-kernel learning algorithm to recognize interlacing point features. The experimental results indicate that the established recognition algorithm can not only effectively recognize the interlacing points in plain,twill and satin weave fabrics,but also has ideal robustness and universality for the adjustment of fabric components and yarn forming process. The average recognition rate achieved in this research reaches 91.2%.

关 键 词：色纺织物 组织点识别 混合色彩空间 多核学习 支持向量机 

分 类 号：TS101.9[轻工技术与工程—纺织工程]