摘要
为了解决纤维横截面法在混纺织物的成分测定中常遇到图像失真、定性效率较低、定量测定软件准确率不稳定、测定效率有待提高等问题,改进了非包埋纤维横截面图像采集方法,测定纤维横截面轮廓特征结构参数,依据圆形度、异形度指标,同时增加了可原位溶解观测的微通道超薄纤维切断装置,建立了混纺织物智能定性模型,对常规纤维进行初步分类,利于快速自动检测。应用YOLOv5模型的图像识别算法,建立混纺织物智能识别模型,提高了纤维横截面图像边缘特征提取准确性和混纺织物定量模型的测定准确性。经实际样品验证,与化学溶解法相比,算法识别96%的样品数据偏差在5%以内,实现了混纺织物成分测定的数字化和智能化。
In order to solve the problems of image distortion,low qualitative efficiency,unstable accuracy of quantitative measurement software,unimproved measurement efficiency and so on,which are often encountered in fiber cross-section method in the composition determination of blended fabric,the method of collecting cross-section image of non-embedded fiber was improved,and the characteristic structure parameters of fiber cross-section profile were determined.According to indexes of circularity and special-shaped degree,a micro-channel ultra-thin fiber cutting device that could achieve in-situ dissolution and observation was added at the same time.An intelligent qualitative model of blended fabric was established for preliminary classification of common fibers.It was beneficial to rapid automatic detection.By using YOLOv5 model image recognition algorithm,an intelligent recognition model for blended fabric was established,which improved the accuracy of edge feature extraction of fiber cross-section image and measurement of quantitative model for blended fabrics.Through practical sample verification,the deviation of the algorithm was within 5%in 96%of the samples compared with chemical dissolution method.Digitization and intelligentize were achieved in component determination of blended fabrics.
作者
秦介垚
卢铭曦
刘小亮
王静
巫莹柱
QIN Jieyao;LU Mingxi;LIU Xiaoiang;WANG Jing;WU Yingzhu(Wuyi University,Jiangmen,529000,China;China Textile Standard(Shenzhen)Testing Co.,Ltd.,Shenzhen,518000,China;Foshan Zhongfanlian Inspection Technology Service Co.,Ltd.,Foshan,528000,China)
出处
《棉纺织技术》
CAS
北大核心
2023年第7期9-14,共6页
Cotton Textile Technology
基金
江门市科技计划项目(2021030103890006796)
揭阳市揭榜挂帅制重大项目(2022DZX027,skjcx033)
创新创业及攀登计划项目(pdjh2021b0514,2022CX43,2022CX44,2022CY61,pdjh2023b0533)。