Vision-based Denim Quality Assessment

Manual color quality control is a slow process in textile industry. The color separation of denim products is judged by human operators to decide the correct chemical process to be applied such as ozone washing, to obtain the desired colors and textural characteristics for end-users. This paper aims at designing a vision-based automated system for grouping industrial denim jeans in order to eliminate human error or any other external factor, and increase efficiency and reliability of the process. Firstly, uniquely identifying features are extracted from images of the products taken in similar conditions using an image capture system, that is specifically designed for it, then parameters of clustering algorithms such as the number of clusters are decided by considering different validity scores. Finally, using the outcomes of several algorithms, a clustering is made from a batch of denim products and then the results are sent to human operators to be evaluated.