Abstract:

In order to solve the problems of the empirical spadework and time consuming when confirming the proportions of the colored fibers in the colored spun yarns, problems associated with the help of the video microscopy and image processing technology, a novel method based on a clustering algorithm is proposed to classify the color fibers and inspect the proportions of different color fibers. Firstly, the colored spun yarns are twisted into fibers, the colored fibers are arranged on the slide under mild tension, and the corresponding images are captured by video microscopy. Secondly, the gray projection method is adopted to localize the colred fibers, and the average value of R, G, B of all pixels in the center line of each colored fiber is extracted as a feature vector to characterize the fiber respectively. Finally, the feature vectors in RGB model are converted to the L*a*b* color model, a coustering algorithm by searching density peaks is applide into classifying the colored fibers and the  proportions of each color fibers are calculated. Experiments results demonstrate that the proposed method can inspect the colors and proportions of colored fibers in the colored spun yarns automatically with a satisfactory accuracy.

Key words: color fibers, color classification, image processing, custering algorithm