纺织学报 ›› 2021, Vol. 42 ›› Issue (02): 101-106.doi: 10.13475/j.fzxb.20201008406

• 纺织工程 • 上一篇    下一篇

基于卷积神经网络的机织物密度均匀性检测

孟朔, 夏旭文, 潘如如(), 周建, 王蕾, 高卫东   

  1. 生态纺织教育部重点实验室(江南大学), 江苏 无锡 214122
  • 收稿日期:2020-10-30 修回日期:2020-11-11 出版日期:2021-02-15 发布日期:2021-02-23
  • 通讯作者: 潘如如
  • 作者简介:孟朔(1996—),男,硕士生。主要研究方向为纺织图像处理。
  • 基金资助:
    国家自然科学基金项目(61976105);江苏省研究生科研与实践创新计划项目(KYCX20_1942);中国纺织工业联合会应用基础研究项目(J202006)

Detection of fabric density uniformity based on convolutional neural network

MENG Shuo, XIA Xuwen, PAN Ruru(), ZHOU Jian, WANG Lei, GAO Weidong   

  1. Key Laboratory of Eco-Textiles (Jiangnan University), Ministry of Education,Wuxi, Jiangsu 214122, China
  • Received:2020-10-30 Revised:2020-11-11 Online:2021-02-15 Published:2021-02-23
  • Contact: PAN Ruru

RichHTML

18

PDF (PC)

155

摘要：

针对目前基于机器视觉的机织物密度自动检测时织物检测视野小、精度低、品种适应性差的问题,提出一种基于多尺度卷积神经网络的检测方法。首先设计了一套离线图像采集系统连续采集织物图像,并建立一个包含详细织物参数的织物图像数据集;然后采用一种具有不同大小局部感受野的多尺度卷积神经网络适应不同大小的织物结构特征,定位纱线位置;最后利用霍夫变换及灰度投影方法处理网络模型所预测的纱线位置图,计算织物经纬密度,并对织物密度均匀性做出评价。结果表明:与其他方法相比,本文方法对于不同类型织物的经纬密度计算误差小于2%,检测精度更高,品种适应性更强。

关键词: 机织物, 织物密度, 织物参数, 图像处理, 卷积神经网络

Abstract:

Aiming at the narrow field of vision, low precision, and low adaptability in the current automatic methods for measuring fabric density, a multi-scale convolutional neural network was proposed. An offline image acquisition system was designed to acquire fabric images continuously, and a fabric image dataset containing detailed fabric parameters was established. Then, a multi-scale convolutional neural network with different sizes of local receptive fields was used for dealing with different fabric structure features and for locating yarns. The Hough transform and gray projection method were used to process the predicted yarn position map in order to calculate the warp and weft density and evaluate the fabric density uniformity. The results show that, compared with other methods, the detecting error of the warp and weft densities for different types of fabrics is less than 2%, which indicates that the proposed method has a higher accuracy and stronger variety adaptability.

Key words: woven fabric, fabric density, fabric structural parameter, image processing, convolutional neural network

中图分类号: 

  • TS101.92

图1

图像采集系统"

图2

数据集经纬密度与织物组织分布"

图3

标记纱线轮廓的织物图与生成的经纱热力图"

图4

网络结构示意图"

图5

部分织物与预测纱线位置图 注:每个图中从左到右依次为图像系统所采集的织物原图、模型所预测的经纱位置图、模型所预测的纬纱位置图。"

表1

图5中织物自动检测与人工检测方法对比"

织物
编号		 自动检测密度/
(根·(10 cm)-1)		 人工检测密度/
(根·(10 cm)-1)		 误差/%
经密 纬密 经密 纬密 经密 纬密
a 456.18 312.20 453 315 0.70 0.89
b 585.16 430.71 591 433 0.99 0.53
c 583.74 492.17 591 492 1.23 0.03
d 510.79 412.60 512 413 0.24 0.10
e 675.08 400.51 669 394 0.91 1.65
f 615.79 400.08 630 394 2.26 1.54

表2

不同参数下算法的效果对比"

参数设置 MAPE/% MSPE/%
经密 纬密 经密 纬密
σ=1 1.52 2.12 2.44 2.78
σ=0.1 1.90 2.49 4.50 4.88
φ=1 2.90 3.68 6.20 6.90
φ=0.1 1.42 1.71 2.04 2.68
φ=0.01 2.41 2.17 5.31 4.92
λ=0 2.52 2.34 5.58 5.89
λ=0.01 1.59 1.67 2.51 3.51
λ=0.001 1.42 1.71 2.04 2.68
λ=0.0001 1.91 2.11 3.95 4.04

表3

不同算法的织物密度检测效果对比"

方法名称 MAPE/% MSPE/%
经密 纬密 经密 纬密
灰度投影 8.44 10.28 15.24 19.20
MSnet 1.55 1.80 3.33 3.31
改进MSnet 1.42 1.71 2.04 2.68
[1] ZHANG Jie, GAO Weidong, PAN Ruru, et al. A backlighting method for accurate inspection of woven fabric density[J]. Industria Textila, 2017,68(1):31-36.
[2] ZHANG Rui, XIN Binjie. An investigation of density measurement method for yarn-dyed woven fabrics based on dual-side fusion technique[J]. Measurement Science & Technology, 2016,27:085403.
[3] XIANG Zhong, CHEN Kaifeng, QIAN Miao, et al. Yarn-dyed woven fabric density measurement method and system based on multi-directional illumination image fusion enhancement technology[J]. Journal of the Textile Institute, 2019(1):1-13.
[4] 姚穆. 纺织材料学[M]. 3版. 北京: 中国纺织出版社, 2010: 222-223.
YAO Mu. Textile materials[M]. 3rd ed. Beijing: China Textile & Apparel Press, 2010: 222-223.
[5] SCHNEIDER D, GLOY Y S, MERHOF D. Vision-based on-loom measurement of yarn densities in woven fabrics[J]. IEEE Transactions on Instrumentation & Measurement, 2015,64:1063-1074.
[6] 王庆涛, 马崇启, 高雨田, 等. 基于频域特征点提取的素色机织物密度识别算法[J]. 纺织学报, 2014,35(4):47-51.
WANG Qingtao, MA Chongqi, GAO Yutian, et al. Identification algorithm of plain woven fabric density via feature point extraction in frequency domain[J]. Journal of Textile Research, 2014,35(4):47-51.
[7] 何峰, 李立轻, 徐建明. 基于自适应小波变换的织物密度测量[J]. 纺织学报, 2007,28(2):32-35.
HE Feng, LI Liqing, XU Jianming. Woven fabric density measure based on adaptive wavelets trans-form[J]. Journal of Textile Research, 2007,28(2):32-35.
[8] 潘如如, 高卫东. 基于图像处理的机织物密度的高精度识别[J]. 纺织学报, 2008,29(11):128-131.
PAN Ruru, GAO Weidong. High-precision identification of woven fabric density via image processing[J]. Journal of Textile Research, 2008,29(11):128-131.
[9] PAN Ruru, GAO Weidong, LIU Jinhong, et al. Automatic inspection of woven fabric density of solid colour fabric density by the Hough transform[J]. Fibres & Textiles in Eastern Europe, 2010,18:81.
[10] LIN J. Applying a co-occurrence matrix to automatic inspection of weaving density for woven fabrics[J]. Textile Research Journal, 2002,72:486-490.
[11] MENG Shuo, PAN Ruru, GAO Weidong, et al. Woven fabric density measurement by using multi-scale convolutional neural networks[J]. IEEE Access, 2019; 7:75810-75821.
[12] SZEGEDY C, LIU W, JIA Y, et al. Going deeper with convolutions[C]// REED S. IEEE Conference on Computer Vision and Pattern Recognition. Boston: arXiv, 2015: 1-9.
[13] RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[C]// BROX T. International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham: Springer, 2015: 234-241.
[14] WANG Z, BOVIK A C, SHEIKH H R, et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Trans Image Process, 2004,13(4):600-612.
[15] OHTSU N. A threshold selection method from gray-level histograms[J]. IEEE Transactions on Systems Man & Cybernetics, 1979,9:62-66.
[16] ZHANG Y T, SUEN C Y. A fast parallel algorithm for thinning digital patterns[J]. Comm Acm, 1984,27:236-239.
[1] 李东洁 郭帅 杨柳. 基于改进图像阈值分割算法的纱线疵点检测[J]. , 2021, 42(03): 0-0.
[2] 王奕文, 罗戎蕾, 康宇哲. 基于卷积神经网络的汉服关键尺寸自动测量[J]. 纺织学报, 2020, 41(12): 124-129.
[3] 刘沐黎, 袁理, 杨亚莉, 刘军平, 龚雪, 鄢煜尘. 色纺机织物组织结构对其呈色特性的影响[J]. 纺织学报, 2020, 41(09): 45-53.
[4] 张祝辉, 张典堂, 钱坤, 徐阳, 陆健. 广角机织物的织造工艺及其偏轴拉伸力学性能[J]. 纺织学报, 2020, 41(08): 27-31.
[5] 武鲜艳, 申屠宝卿, 马倩, 金利民, 张威, 谢胜. 球形弹体冲击下三维正交机织物结构破坏机制有限元分析[J]. 纺织学报, 2020, 41(08): 32-38.
[6] 邵金鑫, 张宝昌, 曹继鹏. 基于图像处理与深度学习方法的棉纤维梳理过程纤维检测识别技术[J]. 纺织学报, 2020, 41(07): 40-46.
[7] 马颜雪, 王世娜, 李毓陵, 温润. 方格立衬结构机织物的一次成形设计实践[J]. 纺织学报, 2020, 41(06): 42-47.
[8] 金诗怡, 周赳. 具有双层效果的提花-印花-剪花织物的设计[J]. 纺织学报, 2020, 41(06): 48-54.
[9] 王泽霞, 陈革, 陈振中. 基于改进卷积神经网络的化纤丝饼表面缺陷识别[J]. 纺织学报, 2020, 41(04): 39-44.
[10] 王晓华, 姚炜铭, 王文杰, 张蕾, 李鹏飞. 基于改进YOLO深度卷积神经网络的缝纫手势检测[J]. 纺织学报, 2020, 41(04): 142-148.
[11] 张铮烨, 辛斌杰, 邓娜, 陈阳, 邢文宇. 基于边界跟踪测量麻纤维横截面参数的算法研究与应用[J]. 纺织学报, 2020, 41(02): 39-43.
[12] 贾小军, 叶利华, 邓洪涛, 刘子豪, 陆锋杰. 基于卷积神经网络的蓝印花布纹样基元分类[J]. 纺织学报, 2020, 41(01): 110-117.
[13] 孙洁, 丁笑君, 杜磊, 李秦曼, 邹奉元. 基于卷积神经网络的织物图像特征提取与检索研究进展[J]. 纺织学报, 2019, 40(12): 146-151.
[14] 王旭, 杜增锋, 王翠娥, 倪庆清, 刘新华. 贯穿正交机织物结构的参数化三维建模[J]. 纺织学报, 2019, 40(11): 57-63.
[15] 贾高鹏, 宋小红, 李莹, 刘晓丹, 潘雪茹. 铜镍金属涂层机织物拉伸过程中电流的响应[J]. 纺织学报, 2019, 40(10): 68-72.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
京ICP备14006648号　 京公网安备11010502044800号
版权所有 © 2021 《纺织学报》编辑部
地址： 北京市朝阳区延静里中街3号主楼6层《纺织学报》杂志社　邮政编码：100025　
电话：010-65017711,65917740　传真：010-65016539　Email:fangzhixuebao@vip.126.com
本系统由北京玛格泰克科技发展有限公司设计开发