基于多视角图像的纱线三维建模及毛羽测量研究  

Study on three-dimensional modeling of yarn and hairiness measurement based onmulti-view images

作　　者：左新焕 李忠健 王蕾[1] 潘如如[1] 高卫东[1] ZUO Xinhuan;LI Zhongjian;WANG Lei;PAN Ruru;GAO Weidong(Key Laboratory of Eco-Textiles,Ministry of Education,Jiangnan University,Wuxi 214122,China;Key Laboratory ofClean Dyeing and Finishing Technology of Zhejiang Province,Shaoxing University,Shaoxing 312000,China)

机构地区：[1]江南大学生态纺织教育部重点实验室,江苏无锡214122 [2]绍兴文理学院浙江省清洁染整技术研究重点实验室,浙江绍兴312000

出　　处：《丝绸》2023年第11期89-95,共7页Journal of Silk

基　　金：国家自然科学基金项目(61802152);纺织之光应用基础研究计划项目(J202109);浙江省基础公益研究计划项目(LGG21F030007);中国博士后科学基金面上资助项目(2020M681736);江南大学研究生科研与实践创新项目(KYCX-23-ZD01);中国纺织工业联合会科技指导性项目(2022009)。

摘　　要：毛羽是评价纱线外观质量的重要参数之一,但现有的二维测量方法无法描述毛羽空间形态,使得测量结果与实际情况存在一定差别。文章介绍了多视角纱线图像采集装置的构建,对采集的多视角图像进行处理并构建纱线三维模型,对毛羽三维点云进行去噪、细化处理,从而实现对毛羽长度的精确测量。实验结果表明,本方法能有效地获取毛羽的三维信息并准确地测量其长度,与USTER TESTER5条干测试仪、ZweigleHL400毛羽测试仪及FZ/T 01086—2020《纺织品纱线毛羽测定方法投影计数法》标准的测试数据进行比较,进一步验证了这一方法的准确性和实用性。The textile and apparel industry plays a pivotal role in China’s import and export trade.Within the entire textile and apparel production chain,yarn production is considered the most critical link,and the quality inspection of yarn is the key measure to ensure the quality of textiles.Hairiness,as one of the core parameters for evaluating the appearance quality of yarn,is still mainly reliant on photoelectric devices for its detection in China.This method has limited detection accuracy,low efficiency,and inconsistent testing standards.Current image-based two-dimensional measurement techniques cannot accurately depict the three-dimensional morphology of hairiness,leading to discrepancies between measurement results and actual conditions.Addressing the challenge that existing hairiness detection techniques cannot comprehensively capture hairiness information,this paper applies three-dimensional reconstruction technology to the field of yarn hairiness detection,paving the way for intelligent and precise measurements in textiles.In this study,we first constructed a multi-view image acquisition system,achieving the collection of five yarn images from different evenly distributed perspectives on a single image.Following this,based on the imaging properties of the multi-view image acquisition system,we calibrated the collected multi-view images.To balance detection efficiency and accuracy,we employed image processing techniques to separately segment the yarn body and hairiness.Then,by using the contour transformation method and leveraging the geometric principles of the optical path structure of the dual-plane mirror single-camera stereo vision system,we restored the actual image and the four virtual image views’relative positions in space,synthesizing a three-dimensional model of the yarn.From the three-dimensional point cloud of the yarn,the three-dimensional point cloud of the hairiness was separately obtained.Noise reduction was then performed on the three-dimensional point cloud of the hairiness based on the pri

关 键 词：多视角图像 数字图像处理 三维建模 点云去噪 三维细化 毛羽测量 

分 类 号：TS103.7[轻工技术与工程—纺织工程]