Yarn-dyed woven fabric density measurement method and system based on multi-directional illumination image fusion enhancement technology

Abstract

The use of machine vision technologies to detect the density of colored yarn-dyed woven fabric has significantly accelerated during the last two decades. Unlike previous studies, this paper proposes an algorithm based on a multi-directional illumination image fusion technology to weaken the color signals in the interlace region of yarn-dyed fabric using the three-dimensional surface structure of the fabric. Four gray-scale images are first sampled using four directional light sources with a square distribution, and the four images are then fused through different discrete wavelet transform methods to enhance the image contrast between the float yarns and their adjacent interstices. A Butterworth filter, Gaussian pyramid, and Hough transform are applied to the fused image sequentially to improve the accuracy of the skew detection such that a gray-scale projection can be carried out along the yarn direction to locate the position of the weft and warp yarns. Finally, the local weighted regression algorithm with an adaptive width factor is adopted for smoothing the projection curve and improving the accuracy of the yarn density detection. For optimization of the proposed method, the effects of the illumination direction angle, image fusion method, fabric color, and weave pattern on the density measurements were investigated. The experimental results show that the proposed method works well and achieves an acceptable level of accuracy regarding the yarn density detection for yarn-dyed fabric.