Automatic pattern segmentation of jacquard warp-knitted fabric based on hybrid image processing methods

Abstract

This paper reports an automatic pattern separation approach for jacquard warp-knitted fabric, which includes bilateral filter, pyramidal wavelet decomposition and improved fuzzy c-means (FCM) clustering. First, jacquard warp-knitted fabric images are captured and digitized by a scanner in gray mode, and then the bilateral filter is adopted to smoothen the fabric textures formed by various lapping movements of jacquard fabric and to reduce the noise appearing in capturing process. Next, multi-scale wavelet decomposition is applied to lessen calculation burden and to shorten computation time. Finally, the modified FCM clustering is proposed, in which the Mercer Kernel function is used to make some features prominent for clustering, and a weight function is proposed to measure the similarity between the data and the clustering center. The experimental results reveal that this hybrid method can achieve fast and accurate pattern segmentation. It is proved that this study is suitable for the pattern separation of jacquard warp-knitted fabric.