FULLY AUTOMATIC APPROACHES FOR CROSSED-POINTS DETECTION IN WOVEN FABRIC RECOGNITION

Abstract

Fabric analysis is done to determine fabric weave patterns used in the weaving process. The analysis of fabric-weave patterns still relied on human inspection until the middle of the 1980s. Since then, a number of studies have been made of automatic pattern recognition of woven fabric. One of the major problems in automatic woven fabric recognition is bow to detect the areas of interlacing warp and weft yams. This problem is termed 'crossed-points detection', the difficult problem. In this work, new and fully automatic methods containing Otsu, Isodata, Li, Mean Minimum, Triangle and Yen thresholding algorithms and based on Fourier image-analysis techniques as well are proposed. The application of these methods to plain woven fabric demonstrates the ability to solve such crossed-points-detection problems. Finally, the algorithms are evaluated visually by superposing the detected grid image on the initial woven-fabric image. Grid images obtained by Otsu and Isodata thresholding methods involve the crossed points better than other methods. Besides, vertical lines that extracted with Li method are more left than lines that detected using other methods. Moreover, results from Mean method is close results of Otsu and Isodata. Similarly, Mean and Minimum method’s result is close to results of Otsu and Isodata. In Triangle and Yen methods, while horizontal lines are extracted, vertical lines are not extracted. All the detected crossed-points are displayed independently and may be saved as file-format images for further processing.

Keywords

• Woven fabric recognition
• fabric weave patterns
• Fourier image analysis
• crossed-points detection
• texture analysis