Record:   Prev Next
作者 Yu, Xiaobo
書名 Micro structure information analysis of woven fabrics
國際標準書號 9780549963370
book jacket
說明 243 p
附註 Source: Dissertation Abstracts International, Volume: 69-12, Section: B, page: 7793
Adviser: George Baciu
Thesis (Ph.D.)--Hong Kong Polytechnic University (Hong Kong), 2008
In this thesis, a fabric structure analysis system is introduced. This system is proposed to deal with some common problems in woven fabric analysis, such as the density measurement, woven pattern extraction and color analysis. Especially, an information encoding method using fabric woven structure is introduced. This system is an integration of modules which are independent and cooperative function units
The foundation of the system is the woven fabric model. Two kinds of fabric models, the space domain Active-Grid-Model (AGM) and the frequency domain model, will be introduced. Fabric models are explicit representations of fabric features and information in either space or frequency domain. AGM is the derivate of point map which is the most used data format in textile industry. AGM supports a generic data interface to different modules inside the system and also the external applications, such as fabric CAD system. The frequency domain model is a statistical model in 2D Fourier spectrum to represent fabric space periodical features. These two fabric models do not exist independently but closely related to each other. The frequency model can be constructed based on the AGM and some specifications of AGM can be obtained by studying features of frequency model. This dependency is the basis of some proposed algorithms, such as the significant points based satin fabric analysis
Based on the fabric models, five function modules are proposed: fabric density measurement and satin fabric analysis using frequency domain (Chapter 4), yarn extraction (Chapter 5), woven pattern extraction using AGM (Chapter 6), yarn-level color clustering (Chapter 7) and weave code (Chapter 8)
Two fabric analysis methods using frequency domain are introduced: fabric density measurement and satin fabric analysis. These two methods are proposed based on three features of frequency model: fabric density principle ( FDP), significant angle and significant points in Fourier spectrum. The pre-requirement to decipher the frequency features are accurate extractions. A set of fabric frequency feature extraction methods are given. These methods guarantee the interested features can be accurately extracted
Yarns are the basic units of fabrics. Detect geometry shapes of yarn are the precondition of many fabric analysis tasks, such as the woven fabric extraction and yarn-level color analysis. Two yarn location methods are introduced. One is the histogram based method and another is the significant-yarn-segment (sys) searching method. Sys searching is more stable than histogram if the fabric samples are not regular and have irregular curves or twists while histogram based method is more robust when the image condition is low
Base on the fabric active grid model (AGM), a module to extract the fabric woven pattern is introduced. This module includes a four-step method to construct the AGM. The construction of AGM makes use of results of other two modules, the fabric density measurement and the yarn locating. The yarn locating method gives the initial guess of the AGM. After the initialization, an AGM automatic adjustment scheme is produced to gather accurate geometry information of the fabric. The cross point classification based on the edge intensities, and the correction based on neighboring and color information, are introduced. The final results show that this module can efficiently extract the fabric woven patterns
The efficiency of yarn-level color clustering (YLCC) depends on four elements: accurate locating and segmentation of yarn regions, color calibration, a proper color space and an efficient color clustering method. Four solutions are proposed to deal with four key problems in YLCC. The yarn locating and AGM automatic adjustment modules give an accurate segmentation of the yarns. The reference white based color normalization is used to calibrate fabric images during the multiple sampling. The iterative segmentations can extend the 1D clustering to 3D clustering in YCbCr space and the iterative mergence can keep the integrations of independent and distinct clusters while segmenting
In past, the decoding (extraction) of fabric woven pattern is the typical task of woven fabric analysis. In this thesis, a new coding technology, the weave code technology, is introduced. This technology provides a solution to embed information into fabrics based on the fabric woven structure. A detailed introduction to the concept of weave code, the basic alphabet mapping scheme and the composition of a complete weave code system, are given. (Abstract shortened by UMI.)
School code: 1170
Host Item Dissertation Abstracts International 69-12B
主題 Textile Technology
0994
Alt Author Hong Kong Polytechnic University (Hong Kong)
Record:   Prev Next