Abstract:

Aiming at the lack of information in yarn detection from two-dimensional image and at the low accuracy of three-dimensional (3-D) yarn evenness, a calibration method for 3-D yarn evenness based on mirrored image was proposed. Four types of compact cotton yarn with different thickness were selected, and the multi-view images of each yarn were collected in one image by a camera. The collected images were calibrated on the xoz plane and xoy plane, while binarization and morphological opening were carried out respectively to obtain clear binary image of yarn evenness. According to the geometric relationship of the mirror imaging system, the 3-D model of yarn was created, and the number and CV value of white dots on each cross section of yarn were calculated. The modeling accuracy of yarn was evaluated by comparing with the yarn diameter and CV value of yarn diameter experimentally measured by Uster TESTER 5. The results show that the correlation coefficient between the number of pixels in each section of the 3-D yarn model and the diameter is more than 0.987, and the difference of CV value between the proposed method in this research and that from Uster testing is less than 2.36%, which proves the feasibility of the calibration method.

Key words: yarn evenness, three-dimensional model, calibration method, mirrored image, image processing