纺织学报 ›› 2020, Vol. 41 ›› Issue (10): 34-40.doi: 10.13475/j.fzxb.20190502707

• 纺织工程 • 上一篇    下一篇

基于神经网络的籽棉颜色分级检测

徐守东(), 冷奕锦, 吴国新   

  1. 安徽财经大学 棉花工程研究所, 安徽 蚌埠 233041
  • 收稿日期:2019-05-14 修回日期:2020-07-14 出版日期:2020-10-15 发布日期:2020-10-27
  • 作者简介:徐守东(1962—),男,高级实验师。主要研究方向为棉花加工与检验。E-mail:mgsxsd@163.com
  • 基金资助:
    安徽高校自然科学研究项目(KJ2018A0439);安徽高校自然科学研究项目(KJ2019A0650);安徽高校自然科学研究项目(KJ2020ZD004)

Research on color grading of seed cotton based on neural network

XU Shoudong(), LENG Yijin, WU Guoxin   

  1. Institute of Cotton Engineering, Anhui University of Finance and Economics, Bengbu, Anhui 233041, China
  • Received:2019-05-14 Revised:2020-07-14 Online:2020-10-15 Published:2020-10-27

摘要:

为解决籽棉颜色分级问题,构造了一个基于L*a*b*颜色空间的色度检测仪,主要由颜色传感器、光源及外围电路构成。针对用于籽棉颜色等级检测2个关键指标(反射率、黄度)输出不稳定问题,采用了4层BP神经网络和5块标准色板进行反复训练,使得校准后的反射率的变异系数小于0.21%,黄度的变异系数小于1.13%。在籽棉颜色等级检测实验中,制作了覆盖12个颜色等级的480个测试样。经过反复实验发现,使用该色度检测仪对1个测试样品,需要均匀分布10个测量点结果的平均值,才能得到稳定的色度测量值。最后,采用神经网络方法,对480个籽棉试样数据进行分析,其中:80%用于训练;20%用于识别。实验结果表明,对12个颜色等级的480个样品进行测试,得到的检测准确率都超过了90%。

关键词: 籽棉, 籽棉颜色分级, 颜色传感器, 人工神经网络

Abstract:

In order to solve the problem of color classifying for seed cotton, a detection device based on L *a*b* color space, which is mainly composed of color sensor, light source and peripheral circuit, was designed. Aiming at the unstable output of reflectance and yellowness, which are very important for color grading, a four layers BP neural network was used and trained repeatedly with 5 standard color boards. After the calibration, the rectified reflectance has coefficient of variation with less than 0.21% and the rectified yellowness has coefficient of variation with less than 1.13%. In the next experiment,480 specimen of seed cotton which cover up 12 color grades is prepared. After repeated experiments, it is found that the average value of 10 measurement points, which are evenly distributed in one specimen, could be used as color value for a test sample. Finally, a neural network was applied to analyze the 480 seed cotton sample data, of which 80% was used for training and 20% for identification. Experiment results show that the detection accuracy is more than 90% for all 12 color grades.

Key words: seed cotton, color grade of seed cotton, color sensor, neural network

中图分类号: 

  • S562

图1

籽棉颜色分级检测仪的整体结构"

图2

光电检测装置原理图"

图3

光电检测装置侧视图"

图4

比色板数据神经网络校正原理"

图5

非线性校正用神经网络"

表1

标准色板反射率与黄色深度的测量值和标准值"

序号 白板 棕色板 黄色板 灰色板 中色板
Rd +b Rd +b Rd +b Rd +b Rd +b
标准值 78.400 4.800 57.200 12.400 70.400 14.600 56.100 2.800 73.100 8.300
实测值1 78.591 4.713 57.019 12.313 70.505 14.673 55.684 2.721 73.233 8.269
实测值2 78.496 4.875 57.028 12.436 70.424 14.638 56.100 2.845 73.245 8.278
实测值3 80.491 4.499 57.407 12.509 70.399 14.612 56.192 2.947 73.346 8.241
实测值4 77.214 4.884 57.275 12.473 70.464 14.665 56.061 2.775 72.951 8.369
实测值5 78.397 4.916 56.828 12.254 70.433 14.643 56.347 2.906 73.248 8.268
实测值6 78.123 5.060 57.259 12.471 70.332 14.549 56.093 2.550 73.084 8.320
实测值7 78.185 4.835 57.034 12.314 70.383 14.564 56.224 2.746 72.975 8.387
实测值8 78.025 4.937 57.229 12.251 70.342 14.556 56.085 2.597 72.786 8.340
实测值9 78.890 4.775 57.199 12.465 70.255 14.513 56.215 2.887 73.223 8.300
实测值10 79.271 4.543 57.264 12.531 70.374 14.567 56.115 2.714 73.307 8.183
实测平均值 78.568 4.804 57.154 12.402 70.391 14.598 56.111 2.769 73.140 8.295
实测标准差 0.869 0.176 0.171 0.107 0.071 0.055 0.174 0.131 0.182 0.061
实测离差 0.168 0.004 -0.046 0.002 -0.009 -0.002 0.011 -0.031 0.040 -0.005
离差百分比/% 0.210 0.070 -0.080 0.010 -0.010 -0.010 0.020 -1.110 0.050 -0.050

表2

样品测量次数与结果之间的关系"

样品 色泽 10次 20次 30次 40次 50次
均值 标准差 均值 标准差 均值 标准差 均值 标准差 均值 标准差
白棉1级 Rd 83.58 1.03 83.81 0.92 83.72 0.92 83.56 0.91 83.51 0.91
+b 8.44 0.69 8.27 0.67 8.46 0.67 8.54 0.62 8.52 0.60
白棉2级 Rd 82.62 0.88 82.38 0.87 81.69 0.88 81.59 0.88 81.71 0.87
+b 8.14 0.29 8.18 0.29 8.40 0.28 8.45 0.28 8.35 0.27
白棉3级 Rd 82.32 1.16 82.36 1.16 81.46 1.16 81.22 1.15 81.26 1.15
+b 7.32 0.78 7.10 0.67 7.36 0.70 7.38 0.66 7.39 0.61
白棉4级 Rd 75.27 2.21 74.59 1.68 74.63 1.37 74.71 1.19 74.79 1.07
+b 8.05 0.50 8.09 0.44 7.97 0.41 7.93 0.38 7.89 0.35
白棉5级 Rd 68.28 0.69 68.96 0.68 69.63 0.68 70.21 0.66 70.60 0.65
+b 7.79 0.54 7.60 0.53 7.46 0.55 7.37 0.54 7.29 0.53
淡点污棉1级 Rd 79.79 1.07 79.54 1.08 79.34 1.07 79.10 1.06 79.10 1.05
+b 9.83 0.18 9.94 0.16 9.93 0.17 9.92 0.15 9.93 0.14
淡点污棉2级 Rd 74.87 0.58 74.75 0.62 74.87 0.58 74.86 0.57 74.84 0.54
+b 10.08 0.27 10.06 0.26 10.05 0.27 10.04 0.25 10.04 0.24
淡点污棉3级 Rd 65.73 0.85 65.89 0.83 66.43 0.85 66.79 0.85 66.82 0.84
+b 8.66 0.06 8.65 0.06 8.68 0.05 8.70 0.05 8.70 0.04
淡黄染棉1级 Rd 79.17 1.58 77.87 1.55 78.08 1.55 78.12 1.50 78.09 1.51
+b 12.21 0.27 12.19 0.27 12.15 0.25 12.15 0.23 12.14 0.22
淡黄染棉2级 Rd 71.68 1.41 72.01 1.19 71.82 1.19 71.94 1.09 71.92 1.02
+b 11.69 0.52 11.61 0.44 11.71 0.45 11.59 0.48 11.60 0.46
淡黄染棉3级 Rd 58.97 0.33 59.32 0.34 59.59 0.32 59.78 0.30 59.77 0.31
+b 10.72 0.17 10.51 0.17 10.26 0.19 10.13 0.17 10.13 0.18
黄染棉1级 Rd 70.27 2.86 70.13 2.77 72.00 2.61 72.06 2.48 72.30 2.37
+b 14.75 0.34 14.54 0.36 14.31 0.35 14.29 0.34 14.28 0.34

图6

实际棉样分类用神经网络结构"

表3

籽棉样品分类准确率和随机识别率"

序号 标准棉样等级 分类准确率 识别率
1 白棉1级 95.0 97.9
2 白棉2级 100.0 92.7
3 白棉3级 92.5 95.8
4 白棉4级 100.0 91.6
5 白棉5级 100.0 100.0
6 淡点污棉1级 100.0 92.7
7 淡点污棉2级 100.0 94.8
8 淡点污棉3级 100.0 90.6
9 淡黄染棉1级 92.5 96.9
10 淡黄染棉2级 100.0 96.9
11 淡黄染棉3级 100.0 100.0
12 黄染棉1级 100.0 100.0
[1] 连素梅, 俸冰, 陆世栋, 等. HX-410型手持式棉花测色仪性能分析[J]. 中国棉花, 2017,44(2):34-36.
doi: 10.11963/issn.1000-632X.201702011
LIAN Sumei, FENG Bing, LU Shidong, et al. Performance analysis of a hand-held cotton color measuring instrument[J]. China Cotton, 2017,44(2):34-36.
doi: 10.11963/issn.1000-632X.201702011
[2] 冯志新, 安浩平, 吴顺丽. 基于RGB颜色模型棉花杂质检测算法[J]. 计算机与现代化, 2013(4):99-102.
FENG Zhixin, AN Haoping, WU Shunli. Cotton impurity rate detection system design based on machine vision[J]. Computer and Modernization, 2013(4):99-102.
[3] 陆永良, 袁裕禄, 李云飞. 基于国际颜色模型(CIELAB)的棉花颜色级测试方法研究[J]. 中国纤检, 2016(3):90-93.
LU Yongliang, YUAN Yulu, LI Yunfei. A study on the cotton color grade test method based on international color model (CIELAB)[J]. China Fiber Inspection, 2016(3):90-93.
[4] 赵瑞浩. 棉花新国标对HVI测试的影响及对策研究[D]. 上海:东华大学, 2014: 5.
ZHAO Ruihao. Research on the effects and countermeasures on testing by HVI of the new national standaRd of cotton[D]. Shanghai:Donghua University, 2014: 5.
[5] 徐守东, 梁后军, 吴国新. 籽棉颜色级检测仪的研制[J]. 中国棉花加工, 2017(6):26-29.
XU Shoudong, LIANG Houjun, WU Guoxin. Research on instrument of seed cotton color classing system[J]. China Cotton Processing, 2017(6):26-29.
[6] 王文颖, 王立永. 影响HVI抽检指标相符率的因素及改善措施[J]. 中国纤检, 2012(4):46-48.
WANG Wenying, WANG Liyong. The impacting F and improving measurement to corresponding rate of HVI checking index[J]. China Fiber Inspection, 2012(4):46-48.
[7] THOMASSON J A. Foreign matter effects on cotton color measurement: determination and correction[J]. Transactions of the ASME, 1993,36(3):663-669.
[8] THOMASSON J A, SHEARER S A, BYLER R K. Image-processing solution to cotton color measurement problems: part I. instrument design and construc-tion[J]. Transactions of the ASAE, 2005, 48(2):421-438.
doi: 10.13031/2013.18307
[9] THOMASSON J A, SHEARER S A, BOYKIN D L. Image-processing solution to cotton color measurement problems: part Ⅱ. instrument test and evaluation[J]. Transactions of the ASAE, 2005,48(2):439-454.
doi: 10.13031/2013.18310
[10] KANG T J, KIM S C. Objective evaluation of the trash and color of raw cotton by image processing and neural network[J]. Textile Research Journal, 2002,72(9):776-782.
doi: 10.1177/004051750207200905
[11] WILKERSON J B, MOODY F H, HART W E, et al. Design and evaluation of a cotton flow rate sensor[J]. Journal of Electronic Packaging, 2001,6(6):1415-1420.
[12] MUNEEM Shahriar, IAN Scott Fleming, HAMED Sari Sarraf, et al. A machine vision system to estimate cotton fiber maturity from longitudinal view using a transfer learning approach[J]. Machine Vision and Applications, 2013,24:1661-1683.
doi: 10.1007/s00138-013-0506-0
[13] 张建华, 祁力钧, 冀荣华, 等. 基于粗糙集和BP神经网络的棉花病害识别[J]. 农业工程学报, 2012,28(7):161-167.
ZHANG Jianhua, QI Lijun, JI Ronghua, et al. Cotton diseases identification based on rough sets and BP neural network[J]. Transactions of The Chinese Society of Agricultural Engineering, 2012,28(7):161-167.
[14] 袁灿, 唐川田, 李文钊. 一种基于神经网络和遗传算法的桥梁加速度传感器优化布置方法[J]. 西华大学学报(自然科学版), 2018,37(2):13-19.
YUAN Can, TANG Chuantian, LI Wenzhao. A optimal placement method of bridge acceleration sensor based on neural network and genetic algorithm[J]. Journal of Xihua University(Natural Science Edition), 2018,37(2):13-19.
[15] 李尚平, 陈曾雄, 周敬辉, 等. BP神经网络对甘蔗宿根切割质量的预测[J]. 农机化研究, 2018,4(10):11-17.
LI Shangping, CHEN Zengxiong, ZHOU Jinghui, et al. Prediction model of cutting quality of the sugarcane ratoon by BP neural network: based on PSO algori-them[J]. Journal of Agricultural Mechanization Research, 2008,4(10):11-17.
[16] LIU J, WANG H, SUN Y, et al. Real-coded quantum-inspired genetic algorithm-based BP neural network algorithm[J]. Mathematical Problems in Engineering, 2015,3(2):1-10.
doi: 10.1155/S1024123X9700046X
[1] 肖琪, 王瑞, 孙红玉, 方纾, 李聃阳. 织物起毛起球机制的理论模型研究进展[J]. 纺织学报, 2020, 41(02): 172-178.
[2] 何晓昀 韦平 张林 邓斌攸 潘云峰 苏真伟. 基于深度学习的籽棉中异性纤维检测方法[J]. 纺织学报, 2018, 39(06): 131-135.
[3] 韦平 张林 刘翔 王冬 苏真伟. 籽棉中异性纤维的双光源成像检测方法[J]. 纺织学报, 2017, 38(04): 32-38.
[4] 杨建国 熊经纬 徐兰 项前. 应用混合种群遗传神经网络的精梳毛纺工艺参数反演模型[J]. 纺织学报, 2016, 37(07): 149-154.
[5] 曹秋玲 朱苏康. 营盘出土籽棉的鉴定与古代西域的棉花利用[J]. 纺织学报, 2013, 34(12): 16-0.
[6] 庹武;陈谦. 基于改进BP神经网络的的西服肩袖造型研究[J]. 纺织学报, 2010, 31(8): 113-116.
[7] 刘茜;王玉亮. ANN与模糊算法在毛精纺织机效率预报中的对比[J]. 纺织学报, 2009, 30(01): 55-59.
[8] 项前;杨建国;程隆棣. 基于支持向量机的纱线质量预测[J]. 纺织学报, 2008, 29(4): 43-46.
[9] 丛芳;赵野军. 基于神经网络的服装制板系统[J]. 纺织学报, 2008, 29(1): 129-132.
[10] 吕志军;项前;殷祥刚;杨建国. 知识表达及其在毛纺织工艺设计中的应用[J]. 纺织学报, 2005, 26(6): 115-117.
[11] 陈雁;李栋高. 判别颜色感觉的人工神经网络方法[J]. 纺织学报, 2005, 26(1): 64-66.
[12] 成玲;万振凯;臧海英. 基于神经网络的织物风格识别系统探讨[J]. 纺织学报, 2002, 23(03): 69-70.
[13] 张瑞林;蒋静坪. 人工神经网络识别丝织物的研究[J]. 纺织学报, 2002, 23(02): 58-60.
[14] 张瑞林. 人工神经网络评价丝织物织物风格研究[J]. 纺织学报, 2001, 22(03): 46-48.
[15] 张瑞林. 人工神经网络控制粗梳毛纺质量的研究[J]. 纺织学报, 2001, 22(01): 41-43.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!