基于边缘和纹理特征融合的蕾丝花型检索方法  

Edge and Texture Features Combination Based Lace Pattern Retrieval Method

作　　者：张聪 李岳阳 罗海驰 蒋高明[1] 丛洪莲 Zhang Cong;Li Yueyang;Luo Haichi;Jiang Gaoming;Cong Honglian(Engineering Research Center of Knitting Technology, Ministry of Education, Jiangnan University, Wuxi,Jiangsu 214122, China;Key Laboratory of Advanced Process Control for Light Industry, Ministry of Education, Jiangnan University,Wuxi, Jiangsu 214122, China)

机构地区：[1]江南大学教育部针织技术工程研究中心,江苏无锡214122 [2]江南大学轻工过程先进控制教育部重点实验室,江苏无锡214122

出　　处：《针织工业》2018年第4期4-7,共4页Knitting Industries

基　　金：国家工信部智能制造综合标准化与新模式应用项目子课题(1064130201160660);江苏省产学研联合创新资金-前瞻性联合研究项目(BY2015019-20;BY2016022-35)

摘　　要：介绍一种采用边缘和纹理特征融合的蕾丝花型检索方法。首先利用Live Wire分割算法提取出1 327个样本花型和530个测试花型,然后分别提取它们的傅立叶描述子、灰度梯度共生矩阵作为边缘特征和纹理特征,最后将测试花型与样本花型特征匹配后按照相似度大小进行排序,并与特征融合后的排序值进行比较,在Matlab中实现蕾丝花型用户界面检索系统。结果表明,与单一的特征检索方法相比,特征融合方法具有更高的识别率,能够满足大数据库下蕾丝花型的检索需求。This paper introduces a kind of retrieval method of lace pattern through edge and texture feature combination. First, 1327 sample patterns and 530 test patterns are extracted by Live Wire segmentation algorithms. Then, the Fourier descriptors and gray-gradient co-occurrence matrix of lace patterns are extracted as the edge fea-tures and texture features. Finally, the feature vector of each test pattern is matched with feature vectors of all sam-ple patterns and all sample patterns are sorted according to the similarity degree, which is compared with the sorted values after the feature combination, and the lace pattern user interface retrieval system is achieved in Matlab. The results show that features combination method has a higher recognition rate compared with single feature retrieval method, which can meet the retrieval needs of lace pattern by large database.

关 键 词：蕾丝花型 LIVE Wire分割算法 傅立叶描述子 灰度梯度共生矩阵 特征融合 检索方法 大数据 

分 类 号：TS186.4[轻工技术与工程—纺织材料与纺织品设计]