Generation of color formulation in a textile product through Artificial Neural Networks

Laura Delia de Jesús Zavala Ortiz, José Antonio Vázquez López, Paloma Teresita Gutiérrez Rosas, Moisés Tapia Esquivas

Abstract


Color is an essential characteristic in quality for many products; to obtain the appropriate color is necessary to determine the colorants proportions required to get an exact coincidence of the color desired. In the textile industry, the development of an appropriate coloration is an essential quality standard, since it generates a direct impact in a product’s level of attraction. Currently in the textile industry, the color’s formulation development depends on a colorist master, that’s why this process is highly subjective, since it depends from the colorist’s experience to get a coincidence of color in function of his appreciation. At the same time, the color’s visual evaluation depends on environmental factors such as: kind of illumination, light incidence, and the background color applied. Due to the subjectiveness and the low standardization of the color’s formulation in the textile industry, it is necessary the development of new techniques that allow to generate more exact color formulations. Artificial neural networks (ANN) are intelligent tools, which can be trained to imitate the colorists’ way of work, allowing the error minimization between the color to develop and a targeted color. With an ANN is possible to create efficient color recipes, in a faster and continuous way within a manufacturing process. To eliminate the color’s visual subjectivity evaluation, the ANN can be fed with precise colorimetric data from the object under study. In the present work the use of a 3-layer back propagation of an ANN is presented, which generates color formulations for textiles products from color coordinates L, a*, b* delivered from a colorimeter. The total measurement coefficient R obtained by the proposed ANN was 0.99776 having an error margin of 0.0016 and 0.0019 in milliliters for the applied colorants.

Introduction: Color evaluation is a quality standard of great importance, which influences the level of acceptance or rejection of many products in the industrial sector (Li, Wang, & Jing, 2015). In the textile industry, one of the most important processes is the appropriate coloring to dye certain type of fabric through the creation of adequate color recipes (Furferi & Carfagni, 2010). Color valuation in the textile industry depends highly in the product’s visual valuation, which is highly subjective, due to the fact that many environmental factors as well as the evaluator training affect the color’s perception. Currently, many methods have been developed for the color perception analysis. Artificial intelligence techniques and tools have been applied currently in them, such as the ANN (Li, Wang, & Jing, 2015).

This article proposes the development of a system to obtain the color recipes, in which a back propagation ANN was applied to generate color formulations that have a small variation based on a stablished color pattern, applying a colorimeter as a measurement instrument, this due to its low cost regarding the spectrophotometer to obtain the color evaluation colorimetric coordinates that allowed the ANN training applied, reducing the subjectivity of color evaluation and formulation, when applying instrumental techniques within the process.

Method: For the present article, having cotton fabric as a base and 2 pigment vats (blue and yellow), 4 colorimetric samples were generated corresponding to 5 proposed formulations, which were replicated 4 times, thus obtaining a total of 20 experimental units, from which a total of 840 colorimetric data (corresponding to the vectors L, a*, b* in the color space CIELAB); this data was analyzed to determine if significant differences were present that could influence in the network’s performance in the color formulation, with this analysis done the data was applied to train a back propagation ANN.

The ANN was analyzed with 5, 10, 15, and 20 neurons in the hidden layer, to determine the best structure that allows for error minimization in the formulation in a fast and adequate way. The choice of network depended on the correlation coefficient value (R-value) which was pretended to be closet to 1. With this process concluded, a proposal for a color formulation was obtained by the outcome from the ANN, which was analyzed to get the mean squared error (MSE); and to verify the ANN performance built according to the accuracy obtained in the color equalization from the formulation created.

Results: Then ANOVA analysis performed showed that there is a significant difference for the sample variable, but this difference didn’t affect the performance from the ANN training on obtaining the precise and adequate color formulations.

The back propagation ANN designed showed a good performance for each analyzed case; having as a base the value R obtained in the training, the best ANN was selected for the purpose of the research; thus, obtaining the adequate color formulations; which were analyzed to get the mean squared error (MSE) value for each colorant applied and verify the selected ANN performance.

Discussion or Conclusion: A back propagation ANN is able to reduce the errors in the formulation of textiles products, in the present article, with the objective to minimize such errors, a back propagation ANN was trained; which has a 3-5-2 structure and showed a good performance for solving this kind of problems, with the ANN training an adequate color formulation was obtained, which showed a mean squared error of 0.0016628 for the blue colorant and 0.0019724 for the yellow colorant, the total correlation coefficient R obtained by the ANN was 0.99776, this value indicates that the created network is able to explain 99.77% of the variation observed in the data applied for its training.


Keywords


Keywords: color; color proportion; neural networks; back propagation neural network; color recipe

References


Almodarresi, E. S., Mokhtari, J., Almodarresi, S. M., Nouri, M., & Shams Nateri, A. (2013). A scanner based neural network technique for color matching of dyed cotton with reactive dye. Fibers and polymers, 14(7), 1196-1202.

Berns, R. S. (2000). Billmeyer and Saltman Principales of color technology. New York, USA: Wiley- Interscience.

Fausett, L. (1994). Fundamentals of neural networks, architectures, algorithms and applications. New jersey, USA: Prentice Hall.

Flórez, L. R., & Fernández, F. J. (2008). Las redes neuronales artificiales, fundamentos teoricos y aplicaciones practicas. La Coruña, Oleiros, España: Netbiblo.

Furferi, R., & Carfagni, M. (2010). An as short as possible mathematical assessment of spectrophotometric color matching. Journal of Applied Sciences, 10(18), 2108-2114.

Furferi, R., Governi, L., & Volpe, Y. (2016). Color matching of fabric blends: Hybrid Kubelka-Munk + artificial neural network based method. Journal of Electronic Imaging, 25(6), 061402-1- 061402-10.

Gorji Kandi, S., Ameri, F., & Khalili, N. (2012). Instrument dependency of Kubelka- Munk theory in computer color matching. Progress in color, colorants and coatings , 85-90.

Gutiérrez-Rosas, P. T., Vázquez-López, J. A., Yáñez Mendiola, J., & López-Juárez, I. (2015). Color's measurement and discrimination, of a cosmetic product by an artificial vision system. 2015 CHILEAN Conference on Electrical, Electronics Engineering, Information and Communication Technologies, 803-808.

Jiang, H. Y., Yang, F., Wang, Z. D., & Deng, L. J. (2007). K/S value prediction of the reactive dyeings with BP nerve network. Dyeing and Finishing, 9, 8.

Li, P. F., Wang, J., & Jing, J.-f. (2015). Application of improved back propagation algorithm in color difference detection of fabric. color research and aplication, 40(3), 311-317.

Mangine, H., Jakes, K., & Noel, C. (2005). A preliminary comparison of CIE color differences to textile color acceptability using average observers. Color research and application, 30(4), 288-294.

Montgomery, D. C., & Runger, G. C. (1996). Probabilidad y Estadística aplicadas a la ingenieria. Mexico DF, Mexico: McGraw- Hill.

Pérez-Alvarez, J., Fernández-López, J., & Rosmini, M. (2004). Chapter 13 Chemical and Physical Aspects of Color in frozen Muscle- Based foods. En Y. Hui, G. I. Legarretta, M. H. LIM, K. Murrell, & W. K. Nid, Handbook of frozen foods (Vol. 133). New York: Marcel Dekker Inc.

Pino, D. R., Gómez, G. A., & de Abajo Martínez, N. (2001). Introduccion a la inteligencia artificial, sistemas expertos redes neuronales artificiales y computación evolutiva. Servicio de publicaciones, Universidad de oviedo.

Sungmin, K., Jong, P. H., & Chang, K. P. (2015). Development of a color matching algorithm for digital transfer textile printing using an artificial neural network and multiple regression. Textile Research Journal, 85(10), 1076-1082.

Weiguo, C., Hui, M., Xiang, Z., & Jiajia, S. (2017). Spectrophotometric prediction of pre-colored fiber blends with a hybrid model based on artificial neural network and Stearns–Noechel model. Textile Research Journal, 87(3), 296-304.

Yang, Y., Ma, H., Yang, Q., Zhang, C., & Shen, J. (2018). Recipe prediction of mélange yarn using modular artificial neural network. Journal of the Textile Institute, 109(5), 629-635.




DOI: https://doi.org/10.21640/ns.v10i21.1507

Refbacks

  • There are currently no refbacks.


Copyright (c) 2018 Nova Scientia

Nova Scientia, year 10, issue 20, May – October 2018, is a biannual journal printed by the Universidad De La Salle Bajío, with its address: Av. Universidad 602, Col. Lomas del Campestre, C. P. 37150, León, Gto. México. Phone: (52) 477 7108500, e-mail: http://nova_scientia.delasalle.edu.mx. Chief editor: Ph.D. Ramiro Rico Martínez. ISSN 2007 - 0705. Copyright for exclusive use No. 04-2008-092518225500/102, Diffusion rights via computer net 04 - 2008 – 121011584800-203 both granted by the Instituto Nacional del Derecho de Autor.

Editor responsible for updating this issue: Direction of Research Department of the Universidad De La Salle Bajío, last updated on May 25th, 2018.