Automatic Recognition Of Woven Fabric Parameters Based On Digital Image Analysis

The woven fabric parameters including the count and density of warp and weft yarns, fabric pattern, and the layout of color yarns of yarn-dyed fabric, are the basic information for the production in the textile enterprise. The count of warp and weft is measured with weighing method, and other parameters are obtained by the inspectors with a textile magnifying glass. Manual analysis is not only time-consuming and lab-intensive, but the results are influenced by the subjective factors of the inspectors. There may be difference between the results of different inspectors. Therefore, the enterprises have a desirable demand for an objective and accurate recognition system, which can obtain the woven fabric parameters with high speed, to satisfy the need of flexible production. It is the purpose of our research to utilize image analysis and artificial techniques to construct an automatic recognition system for the woven fabric parameters except for the count of warp and weft yarn.The paper includes some main contents: the theory of skew rectification for woven fabric image; the automatic inspection of woven fabric density; the automatic recognition of solid color woven fabric pattern; the automatic detection of the layout of color yarns, the color effect and the woven fabric pattern of yarn-dyed fabric. The content for each chapter is briefly introduced as follows:In Chapter 1 the background and significance of the topic selection of the thesis is introduced briefly. The research of the recognition of woven fabric parameters based on digital image analysis home and abroad is summarized. The content and method of current research are introduced. The research methods are divided into different groups while the major achievement and the disadvantages are described simultaneously. The solid color fabric and yarn-dyed fabrics are added into normal white fabrics as the research topics in this thesis. In Chapter 2 the image acquisition and the theory of skew rectification of fabric image are introduced. The demands of the fabric image used in this research are analyzed first and the resolution and the size of the captured fabric image are determined. As the fabric images are captured, the skew rectification is introduced as a pre-processing procedure for the recognition of woven fabric parameters. The technologies of skew rectification for low tightness white fabric, high tightness white fabric, solid color fabric, yarn-dyed fabric are analyzed separately. The experimental results of different methods are given simultaneously. The method used is finally chosen based on the experimental result.In Chapter 3 the automatic inspection of woven fabric density is described. To inspect the woven fabric density, the fabrics are divided into three categories according to the kinds of color yarns: solid color fabrics, single-system-mélange fabrics and double-system-mélange fabrics. The density of the three kinds of fabrics is inspected with different methods automatically.To inspect the density of solid color fabrics, Hough transform is adopted to detect the skew angles of warp and weft yarns. Gray-projection method is then used to locate the yarns and the density is then automatically inspected. To inspect the density of single-system-mélange fabrics, a FCM algorithm is used to classify the colors in the fabric image based on Lab color space. The sub image from the classification results includes only one or two kinds of color yarns. The density of single-system-mélange fabrics can then be inspected from the yarn segmentation results. To inspect the density of double-system-mélange fabrics, color-grads image is firstly proposed to enhance the edge information of the yarns. Gray-projection method and cross correlation coefficient based method are adopted to inspect the density.In Chapter 4 the woven pattern of solid color fabrics is recognized. The yarn segmentation results in the density detection are used to locate the floats first, and then a FCM algorithm is presented to recognize the floats based on these features: the average value and the variance values of the float, the energy and the entropy obtained from gray occurrence matrix. BP neural network and pattern database are used to recognize the woven pattern based on the float recognition results. The experimental proves that both the methods can recognize the fabric pattern correctly.In Chapter 5 the color effect of yarn-dyed fabric is recognized. Firstly, the floats are located with the yarn segmentation result in the density detection of double-system-mélange fabrics. Then, the color features of the float are then extracted to classify the floats with a FCM algorithm. The number of yarn colors is obtained simultaneously with cluster validity analysis. At last, the color pattern can be obtained from the float classification results. The repeat unit of the color pattern—color effect is extracted.In Chapter 6 the layout of color yarns and fabric pattern of yarn-dyed fabric are recognized automatically.The recognition of the layout of color yarns based on fabric image is firstly proposed. A FCM algorithm is used to classify the colors of the fabric image in Lab color space first. The layout of color warps and wefts are obtained after traversing all the sub images and determining the color of the yarns. The repeat unit of the layout of color warps and wefts can be finally obtained with the period extracted method. When the color effect is given, genetic algorithm and logical analysis are used to recognize the layout of color yarns respectively. The experimental proves that the methods can complete the recognition of the layout of color yarns.With the relation between the layout of color yarns and the color effect, the float type in the color effect can be partly determined. The pattern database is utilized to recognize the woven pattern of yarn-dyed fabric finally.In Chapter 7 a summary is made to describe the main contributions and the problems of the present work. The advice of the future work of the recognition of woven fabric parameters is given at last.