DIGITAL IMAGE PROCESSING TECHNIQUES FOR AUTOMATIC TEXTILE QUALITY CONTROL

摘要

This article dealth with a complex problem of textile quality control by the means of computer vision using advanced digital signal and image processing tools. An investigation into a computer vision solution to quality inspection of textiles suggested a texture analysis approach. Since each texture analysis method presents a different potential for analysis of textured textile images, a large number of standard approaches from this category was examined. By considering their advantages and potential for successful flaw detection performance in homogeneous and jacquard textiles, six standard texture analysis techniques were identified for further investigation. A comprehensive examination and evaluation of these chosen techniques resulted. Normalised cross-correlation and SCLC approaches were identified as suitable candidates for real-time flaw detection in textiles with homogeneous structure. Inspection of jacquard textiles, however, presents a problem of greater complexity, the solution to which has not as yet been documented in the literature. To address this problem, a need for the application of a comprehensive cojoint spatial spatial frequency approach was identified. A Gabor filter approach was chosen as a suitable representative of this class of techniques. This research then successfully applied optimised 2-D Gabor filters to the textile flaw detection problem and provided a further support of their suitability for this task. A novel optimised 2-D Gabor algorithm presented in this study is an automatic solution which is adaptable to detect a large variety of textile flaw types, both structural and tonal.