Application of improved back propagation algorithm in color difference detection of fabric

Color is an indispensable indicator of product quality evaluation. To detect the color difference of fabrics, the Levenberg–Marquardt optimized back propagation (BP) algorithm is adopted to extract the color feature values of fabric images. First, RGB values are three inputs of BP neural network, and L*a*b* values measured by spectrophotometer are three outputs of the network. The trained network can obtain the corresponding L*a*b* values conveniently. Then the color difference can be calculated through color difference formula and the characteristic values obtained above. Finally, compared with the color difference calculated by the spectrophotometer, the most appropriate formula can be selected from the four formulas listed in the article (CIEDE2000, CMC, CIE94, and CIELAB) to acquire satisfying results. The experimental results reveal that the color difference of fabrics can be detected with a high accuracy and efficiency with this method. Plenty of duplication workloads and some complex conversion formulas can be avoided, making the acquirement of color difference more efficiently. © 2014 Wiley Periodicals, Inc. Col Res Appl, 40, 311–317, 2015     

Automatic detection of layout of color yarns of yarn‐dyed fabric. Part 1: Single‐system‐mélange color fabrics

To recognize the layout of color yarns of single‐system‐mélange color fabric automatically, a novel FCM‐based stepwise classification method is proposed in this article. This method consists of three main steps: (1) warp yarn segmentation, (2) weft color recognition, and (3) the layout of color warps recognition. In the first step, the yarn segmentation method based on mathematical statistics of subimages is adopted to localize warp yarns preliminarily; and then the segmentation results of warp yarn are corrected by misrecognized‐boundary remove and missing‐boundary interpolation. In the second step, the weft color is extracted based on RGB color histograms of whole fabric image. In the third step, the pixels in each warp yarn are classified into two clusters by fuzzy C‐means clustering (FCM) algorithm in CIELAB color model separately, and the preliminary recognized layout of color warps is obtained. All warp colors are clustered by FCM algorithm in CIELAB color model again and the precise layout of color warps is output. The experimental and theoretical analysis proved that the proposed method can recognize the layout of color yarns of single‐system‐ mélange color fabrics with satisfactory accuracy and good robustness. © 2015 Wiley Periodicals, Inc. Col Res Appl, 40, 626–636, 2015     

Automatic detection of layout of color yarns of yarn‐dyed fabric. Part 3: Double‐system‐Mélange color fabrics

Automat layout detection of color yarns is necessary for weaving and producing processes of yarn‐dyed fabrics. This study presents a novel approach to inspect the layout of color yarns of double‐system‐mélange color fabrics automatically, which is Part III of the series of studies to develop a computer vision‐based system for automatic inspection of color yarn layout for yarn‐dyed fabrics. The inspection of single‐system‐mélange color fabrics has been realized in Part I of the series of studies. Integrating the projection‐based region segmentation method proposed in Part I and the FCM‐based stepwise classification method proposed in Part II, the proposed approach is composed of three steps: (1) fabric region segmentation, (2) fabric region selection, and (3) layout of color yarns recognition. In the first step, the fabric regions are segmented by the projection‐based region segmentation method. In the second step, the reasonable fabric regions are selected by analyzing their color histograms and comparing their weft color's frequency. In the third step, the layout of color yarn is recognized by the FCM‐based stepwise classification method, and the precise layouts of color warps and wefts are produced. The experimental analysis proved that the proposed method can recognize the layout of color yarns of double‐system‐mélange color fabrics correctly by testing four different color fabrics and three pieces of same yarn‐dyed fabrics. © 2016 Wiley Periodicals, Inc. Col Res Appl, 42, 250–260, 2017     

An intelligent skin‐color capture method based on fuzzy C‐means with applications

Consumer behavior is complicated. In the cosmetic market, personal intuition and fashion trends for color selection are guidelines for consumers. A systematic method for female facial skin‐color classification and an application in the makeup market are proposed in this study. In this article, face recognition with a large number of images is first discussed. Then, an innovative method to capture color at selected points is presented and complexion‐aggregated analysis is performed. This innovative method is an extension of face‐recognition theory. Images in RGB format are converted to CIELAB format during data collection and then Fuzzy C‐means theory is used to cluster and group the data. The results are classified and grouped in Lab value and RGB index. Two programs are created. The first program, “FaceRGB,” captures color automatically from images. The second program, “ColorFCM,” clusters and groups the skin‐color information. The results can be used to assist an expert system in the selection of customized colors during makeup and new‐product development.     

Improve neural network‐based color matching of inkjet textile printing by classification with competitive neural network

Nowadays, with increasing use of digital printing in the textile industry, characterization and color matching are very much considered. There is a very complicated relationship between pixel values of input digital image and colorimetric parameters of printed textile samples. One of the most important used methods for inverse characterization of printer and prediction of CMYK digital values is neural network. In this study, the prediction accuracy of CMYK digital values were improved by dividing the training samples into 2, 4, 6, 8, and 10 subgroups using creating a competitive neural network. For classification of samples, L*a*b* or XYZ were introduced to a competitive neural network as input parameters. Then, the classification of test samples was performed by trained competitive neural network. To predict the of CMYK digital values of input digital image, a cascade‐forward back propagation neural network is trained by L*a*b* of each subgroup. The results obtained show that the prediction accuracy of CMYK digital values were improved by suggested method. The best result was obtained by classification of samples with L*a*b* into eight subgroups and using a cascade‐forward back propagation neural network with 4, 4, and 4 neurons in hidden layers.     

Color‐difference formula performance for several datasets of small color differences based on visual uncertainty

Visual uncertainty, while reported, is not used routinely when evaluating color‐difference formula performance in comparison with visual data; rather, data are analyzed assuming no uncertainty; that is, repeating the experiment would result in the identical average results. Previously, Shen and Berns developed three methods to determine whether a color‐difference formula was well‐fitting, under‐fitting, or over‐fitting visual data when visual uncertainty was considered, the method dependent on how the uncertainty was reported and the colorimetric sampling of the color‐difference stimuli. The “nonellipsoid standard error method” was used in the current analyses. Three datasets were evaluated: BFD‐P, Leeds, and Witt. For the BFD‐P data, incorporating visual uncertainty led to the same performance results as the average results, that CIEDE2000 was an improvement over CIE94, which was an improvement over CIELAB. For the Witt data, incorporating visual uncertainty led to the same performance results as the average results, that CIEDE2000 and CIE94 had equivalent performance, both an improvement over CIELAB. However, both formulas under‐fitted the visual results; thus, neither formula was optimal. For the Leeds dataset, the visual uncertainty analysis did not support the improvement of CIEDE2000 over CIE94 that occurred when evaluating the average results. Both formulas well fit the visual data. These analyses also provided insight into the tradeoffs between the number of color‐difference pairs and the number of observations when fitting a local contour of equal perceived color difference: In particular, increasing the number of observations was more important than increasing the number of color‐difference pairs. Finally, average standard error could be used to approximate visual uncertainty defined using STRESS. © 2010 Wiley Periodicals, Inc. Col Res Appl, 2011     

Media‐dependent color appearance modeling based on artificial neural networks

In this study, we tried to consider various color appearance factors and device characterization together by visual experiment to simplify the across‐media color appearance reproduction. Two media, CRT display (soft‐copy) and NCS color atlas (hard‐copy), were used in our study. A total of 506 sample pairs of RGB and HVC, which are the attributes of NCS color chips, were obtained according to psychophysical experiments by matching soft copy and hard copy by a panel of nine observers. In addition, a set of error back‐propagation neural networks was used to realize experimental data generalization. In order to get a more perfect generalizing effect, the whole samples were divided into four parts according to different hues and the conversion between HVC and R_HVCG_HVCB_HVC color space was implemented. The current results show that the displays on the CRT and the color chips can match well. In this way, a CRT‐dependent reproduction modeling based on neural networks was formed, which has strong practicability and can be applied in many aspects. © 2006 Wiley Periodicals, Inc. Col Res Appl, 31, 218–228, 2006; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/col.20209     

A general form of color difference formula based on color discrimination ellipsoid parameters

A general color difference formula has been derived based on the parameters of color discrimination ellipsoids in the CIE 1976 L*a*b* color space. By using different orders of approximation, the general formula resembles the basic forms of the current formulae. The method described in this article suggests a framework for modifying the CIE 1976 L*a*b* color difference formula.     

Crosscultural comparison of color terms and preference of persimmon‐dyed fabric,Galchon

This crosscultural study was aimed at correlating color emotions and preference for persimmon‐dyed cotton fabrics known as Galchon. Cotton fabrics were dyed with persimmon powder, in a range of shades, and in some cases were also iron mordanted. Textile and fashion students from Jeju National University in Korea and North Carolina State University (NCSU) in USA participated in the visual assessment of dyed samples and were asked to scale their visual experience and state their emotion and preference for the terms “Bright,” “Heavy,” “Soft,” “Strong,” “Deep,” and “Like.” Korean observers used “Strong” for iron‐mordanted Galchon, and American observers did not associate “Bright” or “Deep” with weakly dyed fabrics. In addition to the subjective terms described, the color preference for samples was quantified using their CIE colorimetric attributes. For Korean observers, the results indicate a correlation between L^* and “Bright,” whereas for Americans a stronger correlation was obtained against “Soft.” American observers' results also show a relationship between C^* and the term “Warm,” especially for dyeings of Galchon at high concentrations. It was also found that iron mordanting affected responses from both groups but only influenced the color preference of Korean observers. © 2015 Wiley Periodicals, Inc. Col Res Appl, 40, 592–604, 2015     

基于模糊递归神经网络的污泥容积指数预测模型

污泥容积指数（SVI）,一个关键的污泥沉降性能评价指标。针对污水处理过程中污泥膨胀关键水质参数污泥容积指数难以准确在线测量,且实验室取样测量方法时间久、精度低,提出了一种改进型的模糊递归神经网络（HRFNN）用来预测污泥容积指数的变化,通过在网络第三层加入含有内部变量的反馈连接来实现输出信息的反馈。实验结果表明,与其他模糊神经网络相比,该网络的规模小、精度高,处理动态信息的能力明显加强。     

A study of the emotional evaluation models of color patterns based on the adaptive fuzzy system and the neural network

Proposed in this article are two kinds of emotional models based on the neural network and the adaptive fuzzy system that can transform the physical features of a color pattern into its emotional features. The purpose of this system of models is to evaluate the neural network and adaptive fuzzy system for its ability to model psychological experimental data in a way similar to what a human expert would do. Construction of the models was motivated by Soen's psychological experiments, in which he found that such physical features as average hue, saturation, and intensity and the dynamic components of color patterns affected the emotional features represented by a pair of adjectives having opposite meanings. One is based on the neural network in the proposed models, and the other consists of two adaptive fuzzy rule bases and a γ model, a fuzzy set operator, to fuse the evaluation values produced by them. The proposed models showed superior performances compared to Soen's model in the approximation of nonlinear transforms, whereas the latter showed an advantage in obtaining the linguistic interpretation from the trained results. The evaluated results of color patterns can be used to construct a emotion‐based color‐pattern retrieval system, which would be able to recommend the color patterns of a desired human feeling. We believe that in linguistic queries of human feelings, these color‐pattern retrieval systems would be able to select from a gallery the corresponding textile designs, wallpapers, or pictures. © 2002 Wiley Periodicals, Inc. Col Res Appl, 27, 208–216, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/col.10052     

基于混合评价指标的自组织模糊神经网络设计研究

针对在无增长和修剪阈值时模糊神经网络结构难以自适应问题,提出一种基于混合评价指标(hybrid evaluation index, HEI)的结构设计方法。首先,通过模糊C均值聚类算法(fuzzy C-means clustering, FCM)确定初始规则层神经元数目及其中心与宽度。其次,基于戴维森堡丁指数(Davies bouldin index, DBI)和邓恩指数(Dunn index, DI)提出一种新的相关性评价指标(relevance evaluation index, REI)来计算规则层各神经元输出之间的相关性,同时根据训练过程中网络输出均方根误差(root mean square error, RMSE)的变化情况来确定网络的学习能力,然后基于REI和RMSE提出了HEI。通过HEI来调整模糊神经网络的拓扑结构,有效解决了在无增长和修剪阈值时网络结构难以动态自调整的问题且避免了网络结构冗余。最后,通过对Mackey-Glass时间序列预测、非线性系统辨识和大气中PM_2.5浓度预测,证明了该结构设计方法的可行性和有效性。     

基于模糊RBF神经网络的乙烯装置生产能力预测

针对传统的径向基函数(RBF)神经网络隐藏层节点的不确定和初始中心敏感性、收敛速度过慢等问题,提出一种基于模糊C均值的RBF神经网络(FCM-RBF)模型,通过模糊C均值聚类(FCM)得到各聚类中心,基于误差反传的梯度下降法训练隐藏层到输出层之间的权值,克服传统RBF模型对数据中心的敏感性,优化确定RBF神经网络隐藏层的节点数,提高网络训练速度和精度。最后将其用于乙烯装置生产能力预测中,分析预测不同技术、不同规模乙烯装置生产情况,指导乙烯生产,提高生产效率,结果验证了所提出算法的有效性和实用性。     

基于递归模糊神经网络的污水处理控制方法

针对污水处理过程具有非线性、大时变等问题,提出了一种基于递归模糊神经网络的多变量控制方法。该方法通过递归模糊神经网络控制器自适应地获得对操作变量的控制精度,控制器在常规BP学习算法的基础上采用学习率自适应学习算法且引入了动量项来训练网络参数,避免网络陷入局部最优,提高了网络对系统的控制精度。最后,基于仿真基准模型(BSM1)平台对第五分区中的溶解氧和第二分区中的硝态氮控制进行动态仿真实验,结果表明,与PID、前馈神经网络和常规递归神经网络相比,该方法能有效提高系统的自适应控制精度。     

基于遗传算法的模糊神经网络控制器的仿真

对遗传算法 (GA)和模糊神经网络控制器的结构进行了说明。为了克服反向传播算法 (BP)的缺点 ,通过遗传算法对模糊神经网络控制器的参数进行优化 ,亦即对模糊神经网络进行训练。用通过优化后的模糊神经网络控制器控制一个带有纯滞后的非线性对象 ,仿真结果证实了其性能较常规模糊控制器优越。     

Automatic detection of layout of color yarns of yarn‐dyed fabric. Part 2: Region segmentation of double‐system‐Mélange color fabric

To detect the layout of color yarns automatically, a novel projection‐based fabric segmentation method is proposed to segment the double‐system‐mélange color fabric into several regions, which can be seen as single‐system‐mélange color fabrics. This method consists of five main steps: (1) yarn skew detection, (2) fabric image projecting, (3) projection curve smoothing, (4) variance curve calculating, and (5) curve peak confirmation. Based on the acquisition fabric image, the skew angles of warp and weft yarns are detected by Hough transform first. The projection curves of L, a, and b channels in Lab color model are generated and smoothed by Savitzky–Golay filter. The variance curves of L, a, and b are then calculated, and the peaks corresponding to the regional boundaries in each curve are detected. The regional boundaries are confirmed by synthesizing the curve peaks of L, a, and b. The experimental and theoretical analysis proves that the proposed method can segment the double‐system‐mélange color fabric into regions with satisfactory accuracy and good robustness. © 2015 Wiley Periodicals, Inc. Col Res Appl, 41, 626–635, 2016     

基于自适应模糊神经网络的非线性系统模型预测控制

针对非线性动态系统的控制问题,提出了一种基于自适应模糊神经网络(adaptive fuzzy neural network,AFNN)的模型预测控制(model predictive control, MPC)方法。首先,在离线建模阶段,AFNN采用规则自分裂技术产生初始模糊规则,采用改进的自适应LM学习算法优化网络参数;然后,在实时控制过程,AFNN根据系统输出和预测输出之间的误差调整网络参数,从而为MPC提供一个精确的预测模型;进一步,AFNN-MPC利用带有自适应学习率的梯度下降寻优算法求解优化问题,在线获取非线性控制量,并将其作用到动态系统实施控制。此外,给出了AFNN-MPC的收敛性和稳定性证明,以保证其在实际工程中的成功应用。最后,利用数值仿真和双CSTR过程进行实验验证。结果表明,AFNN-MPC能够取得优越的控制性能。     

涤纶纱线和织物的色差影响因素和控制措施探讨

探讨了涤纶纤维染色机理和色差评价方法,初步分析了涤纶纱线、织物在聚酯合成、纺丝以及纺织染整过程中存在的色差影响因素,并提出了控制色差的有效措施。     

Color separation for colored fiber blends based on the fuzzy C‐means cluster

Colored fibers can be blended in a certain proportion to achieve a specific color. It is a very hard task for the colorist to find a good recipe to meet the final product without the aid of computer. In this article, a color separation method for the colored fiber blends is discussed to substitute for some manual work. The fuzzy C‐means cluster is a way to group the color in the colored fiber blends image. The distance index, which is a key factor during the fuzzy C‐means cluster process, is calculated in the RGB color space and the HSV color space with some transformation. The final experiment result proved that the colors of each pixel in the blends' image can be replaced by corresponding cluster center associating colors in the HSV color space, and the main texture as well as the main color information about the fibers in the image is preserved. © 2011 Wiley Periodicals, Inc. Col Res Appl, 2012