Research On Yarn Appearance Detection System Based On Machine Vision

With the rapid development speed of industrialization,people have higher and higher demand on the quality of the textiles,so that more requirements for the speed and efficiency of the detection on yarn quality are put forward.The quality of the yarn not only has an effect on the appearance and the feeling of the textile,but also is closely associated with the sales and price of the product.In recent years,the machine vision technology has provided new ideas for yarn quality inspection.The detection technology based on machine vision can overcome the shortcomings of traditional yarn inspection equipment and provide more accurate parameters to evaluate the yarn quality.The main parts of our research are as follows:(1)The yarn evenness testing based on the machine vision.After the pre-processing of the obtained yarn image,the yarnlevelness is extracted by the maximum entropy-based method,and the diameter of the yarn is obtained by counting the pixels in the vertical direction of the yarnlevelness,and then the yarn evenness is calculated,while at the same time the yarn faults are classified.(2)The detection of the cross-overlapping hairiness.The morphological method is used to reconstruct the image after the yarn image preprocessing,and then the GAC(Geodesic Active Contour)model is employed to check to detect the edge of the yarn hairiness image.Next the corner detection is achieved and the pits are obtained after the false corners are deleted.Finally the dividing lines are obtained by matching the pits to segment the overlapping hairiness.The experimental results show that the proposed method can segment the cross-overlapping hairiness efficiently.(3)The detection of the conventional hairiness.After the yarn image is preprocessed by Laplacian sharpening,the cell-based outlier detection algorithm is used to extract the yarnlevelness profile and the hairiness,and then the improved statistical lines are used to couting the number of yarn hairiness.The advantage of the proposed algorithm is that the yarn hairiness can be directly extracted,which simplifies the algorithm and improves the efficiency.Finally the algorithm is verified and the difference between the results of the proposed method and the USTER results is less than 5%.(4)Configurating the device environment and a simple and user-friendly user interface is designed according to the different functions of the mudules to realizereal-time detection of the yarn appearance and single-frame inspection.Finaly,the research work of the paper is summarized and the future research content of the subject is arranged.A total of 48 frames,18 tables,and 63 references are included in the paper.