DESIGN OF A COMPUTER SYSTEM 
for the 
ANALYSIS OF DEFECTS AND GRADING 
OF WOVEN FABRICS 
by 
SN N i l e s ^ 
This thesis was submitted to the 
Department of Textile & Clothing Technology 
of the University of Moratuwa 
in partial fulfilment of the requirements of the 
1
 "* 
Degree of C V ' ^ \ 
Master of Philosophy 
Department of Textile & Clothing Technology 
University of Moratuwa 
June 2004 u * Tresis call 
SI 6 I 
University of M o r a t u w a 
81618 
8 1 6 1 
I 
D E C L A R A T I O N 
No portion of the work in this thesis has been submitted to any University or 
Institution for any other academic qualification. 
SN Niles 
(Candidate) 
Dr. Sandun Fernando 
(Supervisor) 
(Supervisor) 
1 
ABSTRACT 
Inspection of fabrics is a major consideration in fabric manufacture, as well as in the 
manufacture of garments and other fabric-based goods. In the Sri Lankan industry 
fabric inspection is almost entirely carried out by manual methods, and is therefore 
subjective and prone to human error. This research has sought to address this problem 
by developing a computerised system to analyse and grade fabrics on the basis of 
captured defective images obtained from the fabric. 
In this research a computer-based system for the objective assessment of fabric defects 
was designed. The system was designed with special emphasis on the fabric defects 
occurring in the Sri Lankan industry. Image processing techniques were used to 
analyse scanned images of the test fabric, compare it with an ideal sample which is 
made available, and identify defects according to pre-learnt rules. The information 
gathered was then used to grade the fabric, either by giving the frequency of 
occurrence of defects or by assigning points. 
A new classification method for common defects was designed, that would facilitate 
easy grading according to commonly used grading systems. A coding system for 
defects was also designed, which helps in reporting defects to the user. The detected 
fabric defects were classified and stored according to the developed classification 
method and using the proposed coding system. 
ii 
ACKNOWLEDGEMENTS 
I wish to place on record my sincere thanks to Dr. Sandun Fernando and Dr. Gamini 
Lanerolle, my supervisors, for the invaluable guidance, advice and encouragement accorded to 
me during the period of my research. I would like to thank Dr. Sandun Fernando specially for 
encouraging me to publish, and Dr. Lanerolle for the motivation he gave me during a crucial 
stage of the research. 
I wish to thank the University of Moratuwa for giving me the opportunity for postgraduate 
research. Thanks are due to the Director, Postgraduate Studies, Secretary, Higher Degrees 
Committee, and Dean, Engineering for their encouragement and support. Thanks are also 
conveyed to Prof. Lakdas Fernando and Dr. Nirmali de Silva, former Heads of Department, 
and to Mr. DPD Dissanayake, current Head of Department, for their support and 
encouragement. I specially thank Mr. Dissanayake and Dr. de Silva for willingly serving on 
my progress review committee.'I would also like to thank my colleagues in the Department of 
Textile & Clothing Technology for their support in various ways. 
My project had a considerable computing component, and 1 thank all those who gave me 
valuable input in this matter: Dr. Gihan Dias, former Head of the Department of Computer 
Science and Engineering, Dr. Lalith Gamage, formerly Senior Lecturer of the University of 
Moratuwa, and Dr. N. Kodikara, Head of the Department of Communication and Media 
Technologies of the University of Colombo. 
A very special word of thanks to my friend and colleague, GC de Silva of the Department of 
Computer Science and Engineering, who took time off from his graduate studies and heavy 
work schedules to promptly and painstakingly answer my innumerable questions. My 
gratitude also to a dear friend, Ruwan Fernando, Tech Lead at Virtusa (Pvt.) Ltd. for his 
persistence in keeping me on track on my research and for innumerable pointers regarding 
general computing aspects. 
Thanks are also due to technical officers Ms. Dilum Dissanayake, Mr. HM Seneviratne and 
Mr. GHD Wijesena, Mr. KT Anurasiri, former technical officer, and Mr. HS Soysa, laboratory 
assistant, for their assistance in the preparation of samples for testing, and to Mr. JA 
Chinthaka, laboratory assistant, and Mr. Sanjeewa Silva, office aide, for their assistance in 
printing the thesis. Thanks are also due to personnel in various textile and garment factories, 
especially my former students, for their assistance in data collection and verification. 
Last but not least, I would like to express my gratitude to my parents, for their guidance and 
encouragement throughout my educational career, and to my brother and sister-in-law for 
their support during my undergraduate days. A special word of thanks to my mother, whose 
patience and support has helped me to come this far in the academic field. 
iii 
TABLE OF CONTENTS 
Declaration 
Acknowledgements 
Abstract 
Table of Contents 
List of Figures 
List of Tables 
List of Abbreviations & Acronyms 
Chapter 1 - General Introduction 
1.1. The Importance of Fabric Inspection 
1.2. Objectives of the Research 
1.3. Outline of the Thesis 
Chapter 2 - Literature Review 
2.1. Introduction 
2.2. Early Fabric Inspection 
2.3. Existing Fabric Grading Systems 
2.3.1. 4- Point system 
2.3.2. 10-Point System 
2.3.3. Graniteville ("78") System 
2.3.4. 6-Point System 
2.3.5. BS Standard 
2.3.6. ASTM Standard 
2.3.7. 1 in 9 System 
2.3.8. Comparison of different Grading Systems 
2.4. Fabric Inspection Systems: History & Developments 
2.4.1. Manual Systems 
2.4.2. Semi-automated Systems 
2.4.3. Fully-automated Fabric Inspection 
2.5. Conclusion 
Chapter 3 - Review Of Image Processing & Computer Vision 26 
Techniques 
3.1. Introduction 26 
3.2. Overview of Computer Vision 27 
3.3. Overview of Relevant Image Processing Operations 30 
3.3.1. Image Enhancement 30 
3.3.2. Edge Detection 32 
3.3.3. Morphological Operations 34 
3.3.4. Image Segmentation 36 
3.3.5. Shape Features 38 
3.3.6. Scene Matching 38 
3.3.7. Recognition & Classification 39 
3.4. Conclusion 41 
Chapter 4 - Methodology of the Proposed System 42 
4.1. Introduction 42 
4.2. Fabric Defects 42 
4.3. Relevant Classification of Fabric Defects 43 
4.4. Coding of Fabric Defects 48 
4.5. Introduction to the new system 51 
4.6. Sample Preparation 52 
4.7. Steps of the Algorithm 52 
4.8. An Alternate Approach 68 
4.9. Software used 69 
4.10. Program Details 71 
4.11. Conclusion 74 
Chapter 5 - Discussion & Conclusions 75 
5.1. Introduction 75 
5.2. Classification & Coding System 75 
Chapter 6 - Suggestions for Future Work 79 
6.1. Introduction 79 
v 
6.2. Detection of Colour Defects 
6.3. Extension of system to cover full length and width of fabrics 
6.4. Further Classification 
6.5. Software enhancement 
6.6. System enhancement 
6.7. Interfacing 
6.8. Broadbasing of the scope of the system 
List of References 
Appendices 
Appendix A1 - Defects listed in ISO 8498: 1990 (E/F) 
Appendix A2 - Defects listed in ASTM D3990: 1990 
Appendix A3 - Defects & Imperfections listed in ASQC 4-Poi 
System 
Appendix A4 - Matlab Source Code 
Appendix A5 - Visual Basic Source Code 
Appendix A6 - Program Results 
LIST OF FIGURES 
Fig. 2-1 Manual Inspection in Progress 15 
Fig. 2-2 Detection of Fabric Faults in Manual Inspection 18 
Fig. 2-3 Lindley Narrow Fabric Inspection System 20 
Fig. 2-4 Sectional View of the Uster Fabriscan 21 
Fig. 2-5 I-Tex 2000 system for the inspection of finished fabrics 22 
Fig. 3-1 Diagram of a typical Computer Vision System 29 
Fig. 3-2 Simple Rule-based Classifier 41 
Fig. 4-1 Comparison of the appearance of (a) knot & (b) slub 46 
Fig. 4-2 Comparison of the appearance of different defects 46 
Fig. 4-3 Proposed Coding System 50 
Fig. 4-4 Grey scale image of a fabric scanned on a flatbed scanner 53 
Fig. 4-5 (a-b) Histogram Equalisation examples 54 
Fig. 4-6 (a-f) Edge Detection examples 55-57 
Fig. 4-7 (a-d) Smoothing examples 58-60 
Fig. 4-8 Wiener filter followed by 2-D Convolution using an 61 
averaging filter 
Fig. 4-9 Thresholded image, showing the defective region clearly 63 
Fig. 4-10 Final image after 3-step morphological operation 64 
Fig. 4-11 Block Diagram of the Process 67 
Fig. 4-12 (a-e) Examples of Image Subtraction 68-69 
Fig. 4-13 User Interface of the System 71 
Fig. 4-14 Flow chart of the software program 73 
4 vii 
LIST OF TABLES 
Table 2-1 4-Point System 7 
Table 2-2 4-Point System (KTA) for Tricot Fabrics 8 
Table 2-3 Determination of First Quality of Knitted Fabrics 8 
Table 2-4 10-Point System 9 
Table 2-5 6-Point System 10 
Table 2-6 Comparison of points allocated under different inspection 
systems 
13 
Table 3-1 Common Gradient Operators 33 
Table 4-1 Grouping of observed Fabric defects 45 
Table 4-2 Defect Classification 47 
Table 4-3 List of defects in each category grouped according to the 
new classification 
47 
Table 4-4 Coding of Fabric defects 49 
• V 
4 
viii 
LIST OF ABBREVIATIONS & ACRONYMS 
AFIS Automatic Fabric Inspection System 
ASQC American Society for Quality Control 
ASTM American Society for Testing Materials 
BS British Standards 
DOG Difference of Gaussian filter 
EVS Elbit Vision Systems 
FDAS Fabric Defect Analysis System 
FFT Fast Fourier Transform 
FWA Fuzzy Wavelet Analysis 
ITMA International Textile Machinery Association 
KTA Knitted Textile Association 
LOG Laplacian of Gaussian filter 
UNI DO United Nations Industrial Development 
Organization 
WIRA Wool Industry Research Association