TEKSTIL VE KONFEKSIYON, vol.27, no.2, pp.108-116, 2017 (SCI-Expanded)
In this study, an approach based on Automatic Multiscale-Based Peak Detection (AMPD) is proposed for detecting yarn locations in a woven fabric image with skew and noise. Finding yarn locations in a digital image of woven fabric can be defined as a problem of local maxima detection in a noisy signal of a periodic or a quasi-periodic signal. In this approach, projection profiles of woven fabric images in vertical and horizontal directions are calculated and local peaks are detected by using the AMPD algorithm where local maxima indicate yarn center positions along the woven fabric width or length. The method is tested with real woven fabric images and experimental results show that AMPD algorithm performs well for detecting yarn locations. Although the proposed method performs well on noisy images; detection accuracy is highly affected by the skewness. Therefore, woven fabric images are needed to be enhanced before obtaining signals.