基于SVM方法的色纺花式纱面料自动分类  

Automatic Classification of Colored Spun Fancy Yarn Fabric Based on SVM Method

作　　者：何孟永 潘如如 张宁 高卫东 HE Mengyong;PAN Ruru;ZHANG Ning;GAO Weidong(Intelligent Textile Laboratory,Jiangnan University,Wuxi 214122,China)

机构地区：[1]江南大学智能纺织实验室,江苏无锡214122

出　　处：《东华大学学报（自然科学版）》2020年第4期542-548,共7页Journal of Donghua University(Natural Science)

基　　金：国家自然科学基金资助项目(61501209);中央高校基本科研业务费专项资金资助项目(JUSRP51907A)。

摘　　要：为解决色纺花式纱面料人工分类准确率低的问题,提出一种基于局部二值模式(local binary pattern,LBP)、小波分析和支持向量机(support vector machine,SVM)的色纺花式纱面料自动分类方法。采集色纺花式纱面料图像组成原始图像库;通过对图像数据进行翻转、对比度增强及添加噪声等操作增广图像数据;对增广后的图像数据进行预处理,通过直方图均衡化增加图像背景与前景的对比度,利用LBP等价模式提取图像的局部特征,获得59维特征参数;利用Haar小波对图像进行4层分解,提取每层分解后的低频分量,以及高频分量的水平、垂直、斜线分量,求其标准差后得到16维特征参数。最后将提取到两种特征串联组合为75维特征参数后,输入SVM分类器进行训练和分类。选取增广后的图像数据集进行试验,结果显示,LBP、小波分析与SVM分类器相结合方法的平均分类准确率达到97.0%。To solve the problem of low manual classification accuracy of colored spun fancy yarn fabric,an automatic classification method based on local binary pattern(LBP),wavelet analysis and support vector machine(SVM)classifier was proposed.Colored spun fancy yarn fabric images were captured to form original image database.Several image augmentation operations including image inversion,contrast enhancement and noise addition were carried out to expand the image data.Histogram equalization was used to preprocess the image to increase the contrast.LBP equivalent mode was used to extract local features of the image,and 59 feature parameters were obtained.Haar wavelet was used to decompose the image in four layers.The low-frequency components and the horizontal,vertical and diagonal components of the high-frequency components were extracted after each layer decomposition.The standard deviations were calculated to obtain 16 feature parameters.Finally,the extracted two types of features were combined in series with into a 75-dimension feature,and it was input into SVM classifier for training and classification.The experimental results show that the average classification accuracy rate of the proposed method combining local LBP,wavelet analysis and SVM classifier achieves 97.0%.

关 键 词：色纺花式纱面料 自动分类 局部二值模式 小波变换 支持向量机 

分 类 号：TS941.26[轻工技术与工程—服装设计与工程]