一种基于残差网络的相似动物纤维检测方法  

An identification method of similar animal fibers based on residual network

作　　者：肖玉麟 路凯[1,2] 张洋 刘晔 江超[1] 田润强 Xiao Yulin;Lu Kai;Zhang Yang;Liu Ye;Jiang Chao;Tian Runqiang(College of Information Engineering,Xuchang University,Xuchang 461000,China;Henan International United Laboratory of Polarized Sensing and Intelligent Signal Processing,Xuchang 461000,China)

机构地区：[1]许昌学院信息工程学院,许昌461000 [2]河南省偏振感知与智能信号处理国际联合实验室,许昌461000

出　　处：《现代计算机》2023年第18期35-39,共5页Modern Computer

基　　金：国家自然科学基金(62101478);河南省重点研发与推广专项(科技攻关)项目(212102210402、202102310236、222102320186);河南省高等学校重点科研项目(23B520016);河南省高校大学生创新创业训练计划项目(202210480037)。

摘　　要：羊绒和羊毛纤维的表面形态及物理化学特征非常相似,这两种动物纤维的鉴别一直是纺织领域的难题。提出了一种基于残差网络的纤维检测方法,能够快速准确地识别羊绒和羊毛纤维。实验中首先将采集到的纤维图像使用了图像翻转等数据增强方法,接下来使用残差网络训练和测试数据样本,测试集的平均识别率达到了96.56%。The surface morphology and physical and chemical characteristics of cashmere and wool fibers are very similar.The identification of these two animal fibers has always been a difficult problem in the textile field.An identification method of fibers based on a residual network is proposed,which can quickly and accurately identify cashmere and wool fiber.In the experiment,the collected fiber images were first used with data enhancement methods such as image flipping,and then the residual network was used to train and test data samples.The average recognition rate of the test set reached 96.56%.

关 键 词：羊绒 羊毛 识别 显微镜图像 残差网络 

分 类 号：TS107.2[轻工技术与工程—纺织工程] TP391.41[轻工技术与工程—纺织科学与工程]