Tae Jin Kang

Automatic Recognition of Fabric Weave Patterns by Digital Image Analysis

By using computer image processing and analysis, a system to detect both weave patterns and yarn color designs is developed in this study. The image of a woven fabric is captured by a Hitachi color CCD camera and converted into digital data by a Targa+32 board. Two images—transmitted and reflected—are used to detect weave patterns. From the transmitted images, warp and weft cross points and the sizes of the yarns are determined by analyzing gray value changes in both horizontal and vertical directions. Then the warp and weft crossed states are determined by analyzing the normalized aspect ratio of an ellipse-shaped image at crossed points of the fabric. Furthermore, from reflected images, the total number of yarn colors and their arrangements in the fabric are determined. An HSV color model differentiates or groups similar yarn colors. Consequently, the system allows the weave pattern, either colored or solid, and the color design of a fabric to be correctly recognized.By using computer image processing and analysis, a system to detect both weave patterns and yarn color designs is developed in this study. The image of a woven fabric is captured by a Hitachi color CCD camera and converted into digital data by a Targa+32 board. Two images—transmitted and reflected—are used to detect weave patterns. From the transmitted images, warp and weft cross points and the sizes of the yarns are determined by analyzing gray value changes in both horizontal and vertical directions. Then the warp and weft crossed states are determined by analyzing the normalized aspect ratio of an ellipse-shaped image at crossed points of the fabric. Furthermore, from reflected images, the total number of yarn colors and their arrangements in the fabric are determined. An HSV color model differentiates or groups similar yarn colors. Consequently, the system allows the weave pattern, either colored or solid, and the color design of a fabric to be correctly recognized.

Garment pattern generation from body scan data

An automatic garment pattern design system using three-dimensional body scan data has been developed. A body model has been generated from massive body scan data using segmentation and the Fourier series expansion method. The surface geometry of a standard garment model used in the apparel industry was reconstructed by stereovision technique and converted into a mesh structure. Surface warping algorithm was used to make an equalized geometry of two models, and multi-resolution mesh generation along with optimum planar pattern mapping algorithm were used to generate the optimum two-dimensional patterns of the garment.

Automatic Structure Analysis and Objective Evaluation of Woven Fabric Using Image Analysis

An automatic fabric evaluation system has been developed to automatically analyze the structure of woven fabric and objectively evaluate fabric quality. Fabric images are captured by a CCD camera and preprocessed by Gaussian filtering and histogram equalization. Fabric construction parameters such as count, cloth cover, yarn crimp, fabric thickness, and weight per unit area are measured automatically from planar and cross-sectional images of woven fabric with image processing and image analysis. Results obtained with the system show good correspondence with experimental values. In order to evaluate the quality of woven fabric, defects such as slubs or missing picks are detected successfully from defect images, and the uniformity of yarn spacing and orthogonality of the yarn intersecting angle are determined from normal fabric images. The coefficients of variation of yarn spacing and the yarn intersecting angle are measured quantitatively so that quality can be compared using these values.

Development of three‐dimensional apparel CAD system: Part 1: flat garment pattern drafting system

A comprehensive apparel CAD system was developed to perform automatic garment pattern drafting and the prediction of the final drape shape of designed garment putting on the human body. Three dimensional apparel CAD system starts with a flat garment pattern drafting system. A computerized pattern design script language has been created based on the traditional patterner’s principles to develop an automatic draft system of performing basic garment pattern drafting as well as grading rule generation. A pattern modification system was also developed considering functions required in apparel CAD such as auxiliary pattern generation, seam line creation, and dart manipulation to generate engineering patterns which can be used in the three dimensional garment shape prediction system presented later in part II of this paper.

Development of three‐dimensional apparel CAD system: Part II: prediction of garment drape shape

A fast response three dimensional garment drape shape prediction system has been developed. A human body model generator has been established for the garment draping on it. For the mass production of different size of garments for the various sized body models, the cross‐sectional value from the anthropometric data was used as the standard for the size accommodation to make a resizable human body model. To construct the cloth drape shape prediction system, the finite element analysis method has been utilized. The designed garment pieces were divided into fine quadrilateral elements using a specially coded mesh generating program, then some appropriate sewing conditions were assigned to transform two dimensional patterns into three dimensional shapes. The final drape shape of the garment was determined from the solutions of the contact condition with human body, deformations, and the weights of the elements constituting the garment pieces, as well as the surface texture of the cloth.

Image Analysis of Standard Pilling Photographs Using Wavelet Reconstruction

A new approach to pilling evaluation based on the wavelet reconstruction scheme using undecimated discrete wavelet transform (UDWT), which is shift-invariant and redundant, was investigated. A method of digital image analysis to attenuate the repetitive patterns of fabric surface and enhance the pills is presented. A preliminary evaluation of the proposed method was conducted to SM50 European standard pilling images. The results show that reconstructed resolution level, wavelet bases and subimage used for reconstruction can affect the segmentation of pills and thus pilling grading. The area ratio of pills to total image was effective as a pilling rating factor. The results suggest that the proposed method in this paper is applicable to the practical evaluation of pilling.

Interactive garment pattern design using virtual scissoring method

Purpose – The designing and initial alignments of 2D garment patterns in 3D space are the key procedures in 3D apparel design. This paper presents a new methodology to prepare and edit initial pattern shape in 3D space by simulating virtual cloth scissoring.Design/methodology/approach – In conventional apparel CAD tools, flat 2D patterns are drawn and sewn in 3D space. Thus, the final appearance of 3D garment cannot be easily predictable for non‐specialized personnel from the flat patterns. This paper adopts the real pattern designing method of “draping”, incorporating it into computer‐based designing so that the user can realistically cut, sew and add the cloth by only using a mouse. 2D and 3D meshes are edited simultaneously and thus a flattening process is not needed.Findings – Several mesh‐based operations such as cutting, sewing, adding, and fixing are devised and have been successfully applied to virtual garment cutting.Practical implications – Our new pattern drawing method has an advantage that de...

New Objective Evaluation of Fabric Smoothness Appearance

A new evaluation system is presented for measuring the smoothness appearance of fabric surfaces objectively and quantitatively. In this system, the contour of the fabric surface is measured with the stereo vision algorithm, and the data are then used to evaluate fabric smoothness by fractal geometry, which explicitly explains the degree of ruggedness of the fabric surface as a decimal fraction with precise grading. This study illustrates the stereo vision technique and its image processing for 3D measurements of surface contours using AATCC Test Method 124. The fractal dimensions of replicas are obtained by a fractal geometry algorithm such as reticular cell counting or cube counting. A new equation is established from a linear regression between the fractal dimensions of replicas and their grades. The experimental results show that the new grading based on 3D vision and the fractal dimension corresponds to a visual assessment of fabric smoothness with more accuracy and reliability. The new equation based on the fractal dimension should determine an objective rating of fabric smoothness that can substitute for the conventional subjective AATCC rating method for fabric smoothness and provide a quantitative reliable value to assess fabric smoothness with more accuracy and reproducibility.

Objective Evaluation of Fabric Pilling Using Stereovision

A noncontact three-dimensional measurement method for the objective evaluation of fabric pilling is developed. Three parameters, including the number, area, and density of pills, are defined by analyzing the five conventional standard pilling photographs that characterize pilling. The relationship between pilling grades and those three parameters can be obtained through a multivariable linear regression method. A CCD camera is used to capture the image of a laser line projected on the surface of the fabric specimen, and a simple trigonometric calculation is performed to reconstruct the three-dimensional shape of the fabric. Also, two CCD cameras are used to reconstruct the fabric surface with pills with a stereovision technique. The three-dimensional surface model is transformed into a binary image using a height-threshold algorithm, and the parameters extracted from that image are used to calculate the pilling grade of the specimen.

Optimized garment pattern generation based on three‐dimensional anthropometric measurement

An automatic garment pattern generation system has been developed for the three‐dimensional apparel CAD system. To substitute the garment fitting process, which requires lots of trial and error in the traditional pattern generation methods, we developed a new direct pattern generation method using body‐garment shape matching process. In this method, we first generated a body model using three‐dimensionally measured anthropometric data and transformed it to have a convex shape similar to that of a commonly used dummy model in garment design process. Then a typical garment model has been defined by measuring the surface information of a dummy model using stereoscopy and adjusting its shape considering the geometrical constraints of the underlying body model to obtain the optimum fit garment patterns. Finally, we developed a pattern flattening algorithm that flattens the three‐dimensionally adjusted garment model into two‐dimensional patterns considering the anisotropic properties of the fabric to be used.