Abstract:

In this paper, a novel colour clustering method based on the K-means clustering algorithm is developed for interlaced multi-coloured dyed yarn woven fabrics which can be used to sort the colour of the dyed yarn for the development of a quick response fabric system. Firstly fabric images captured by a flat scanner could be decomposed into three sub-images in red, green and blue channels, respectively. Secondly median filters with different template sizes were selected to process the sub-images in the three color channels separately. Thirdly filtered images in the RGB colour space, reconstructed from the three sub-images, can be converted into the Lab colour format. Ultimately the results of colour segmentation and classification can be obtained based on the Lab color space using the improved Kmeans clustering algorithms. Our experimental results indicated that our method proposed works better than the conventional method based on subjective and manual operations with the aid of simple tools in terms of both accuracy and robustness.