Research On Color Representation Model Of Colored Textile Fabric Based On Two-dimensional Texture Features And Its Application

Textile industry is the traditional pillar of China's national economy and the most important area of people's livelihood,but also the industrial field with the most prominent competitive advantage in China.Because of its delicate and soft appearance,it is favored by consumer groups for its special hazy three-dimensional effect and quality and has formed a unique color spinning product.Compared with the traditional "strip dyeing" and "piece dyeing" process,the color spinning products adopt a new processing process of "dying first,then spinning,then weaving",and their color expression methods are flexible,diverse,and environmentally friendly.The products have strong product competitiveness and high added value.However,in the process of color spinning,whether it is the mixture of dye-dyed fiber or the colored spun yarn or fabric made from the processing of textile,its color characteristics do not completely conform to the "additive color characteristics" and "subtractive color characteristics" because of the color principle of both "physical color mixing" and "chemical color mixing" process.Due to the lack of complete color representation model guidance,in actual production,enterprises need to achieve the target color according to customer samples,rely on experience and repeated trial spinning,visual inspection.This process is not only inefficient,but also prone to introduce a random error of subjective judgment,which affects quality control and commercial delivery,and can not meet the market demand of small-batches,multiple varieties,and fast delivery.Due to the complexity and particularity of the color spinning process,this paper focuses on the construction of the color representation model of color textile.The specific contents are as follows:(1)Given the complexity and changeful state of stacking and aggregation of colored textile fibers,a color representation model integrating 2D local binary model was proposed in this paper.In HSV and Lab mixed color space,the model fuses feature components with similar attributes.At the same time,the pseudo-gray image can be extracted by the 2DLBP(2D Local Binary Pattern)operator,and the co-occurrence matrix features of the acquisition can be improved.On this basis,the color representation model of color textile can be established by using the characteristics of fused third-order color moments.The experimental results show that the constructed color expression model can effectively characterize 33 samples of color textiles with different color schemes.The mixing characteristics are significantly correlated with the sample mass ratio,and the average correlation coefficient is0.23.(2)Aiming at the problem that the 2D local binary mode color representation model has limited ability to color textile texture representation,a texture feature extraction algorithm based on 2D ternary structure descriptor is established.Based on the three-color Spaces of HSV,Lab and YUV,the algorithm obtains ternary Structure texture features and global color features from brightness values and color channels of each color space,and forms one-dimensional uniform ternary Structure descriptors(TSD).Then,the 2DTSD(2D Three Structure Descriptor)features are constructed by using the rotational invariant one-dimensional uniform ternary Structure Descriptor feature graph,to realize the fusion of the 2DTSD features.At the same time,to make a scientific quantitative analysis of the validity of the model,this paper also uses the Pearson correlation coefficient method for experimental analysis.The measured results show that the normalized difference degree of fusion characteristics between samples is positively correlated with the different degree of mass ratio between samples.The overall correlation coefficient is 0.019,and the correlation coefficient variance is 0.0004.Therefore,the two-dimensional ternary structure descriptor is more accurate than the color representation model constructed by the two-dimensional local binary model.(3)Given the difficulty of parameter feature extraction in the process of color textile organization identification,the color representation model based on 2D local binary mode and the color representation model based on 2D terpolymer structure descriptor established in this paper are deepened.Based on the BP network and naive Bayes classification theory,hierarchical hybrid classifier and correct classification(CCPS)are constructed to quantitatively analyze the above two models respectively.The experimental results show that,when the twist coefficient is 330,the recognition rate of plain weave and twill weave fabric is100%,and the recognition rate of satin weave fabric is 92% and 98.8%,respectively.By analyzing the experimental results of the two methods,it is concluded that the recognition rate of 2D ternary structure descriptors is higher.The research and application of this paper are of great guiding significance to the color characterization of colored textile products in enterprises.At the same time,the two fabric organization identification methods established can accurately identify the complex organizational structure of colored textile products,which lays a foundation for the combination of textile industry and information technology engineering and digital transformation.