Yarn Fineness Measurement Based On Image Processing

So far, the detection methods of yarn fineness include weighing test, board regularity inspection visually and projection microscope method. But the real fineness information can't be revealed by the mentioned methods because of its faults, such as time composition, subjective and environmental factors. So, the research on the more efficient and accurate measurement method for manufacture and inspection in textile factory has great significance.Allusion to the existed methods, this paper explores the image processing for the yarn fineness inspection to achieve fast and accurate yarn fineness measurement. At first, yarn images are captured by Motic video microscope, and then based on Matlab R2010b, a system used for fineness measurement including images capturing,image pretreatment, images segmentation and yarn linear density measurement and analyzing is proposed.To get the desired yarn image that has obvious background and target, a yarn image capture system with automatic yarn guide and Motic portable video microscope was proposed. During the preprocessing, the comparison among median filtering, mean value filtering and two-dimensional adaptive wiener filtering has been implemented to decide the optimal filtering method and its filter parameters. Research showed that the filtered images treated by two-dimensional adaptive wiener filtering algorithm with 45 pixels×45 pixels square filtering scale could both filter the interference information including noise and yarn hairiness information and retained more completely the information of yarns'contour and details in images.To segment the yarn image, the theory of OTSU threshold segmentation algorithm is used to get yarn evenness binary image, and then using round structure with 4 pixels radius did open operation to get complete and accurate yarn evenness binary images. With 21s, 32s, 40s, 50s, 60s and 80s pure cotton yarn as test materials, yarn diameter's variation between the same yarn samples with five hundred pictures were measured and studied by the method proposed by this paper. Their probability density functions were obtained by Gaussian fitting to quantify the evaluation of yarn fineness.Finally, the difference and the corresponding reasons between the method by using image processing and weighing method are analyzed from measuring principle and the algorithm. The final measurement results showed that compared with traditional yarn fineness measurement methods, the yarn real fineness and the probability density distribution can be calculated accurately based on image processing technologies more efficiently with less environmental interference.