Studies On Image Measurement Of Fabric Weave Parameters

In this paper,we analyze development and defects of image processing techniques and provide some recognition algorithms to detect the structural parameter in woven fabrics,such as warp and weft densities,yarn diameter,linear density of the yarn and yarn twist angles using Digital Image Processing and realize automatically function of system with mixed program VB and MATLAB.Previously,there are many studies in theory,but studies in practice is less.In this paper,a practicable system is developed to detect warp and weft densities in woven fabrics. As simulated a textile analysis magnifying glass to design algorithms,the system can count automatically the number of warp and wept yarns by recognizing the yarn gaps to realize automatic measurement of warp and weft densities.Finally,this paper provides the reference test program of 2 centimeter detected fragment for satisfying the small samples. Experimental results demonstrate that the most of conventional fabrics,such as plain fabric, printed fabric and colored weaving fabric,can be detected using this method.Compared to the results of man-measurement,relative errors of image measurement system are within 2%.So the system we designed has higher application value for detecting warp and weft densities in woven fabrics.Because of thickness unevenness,non-circular cross section,yarn hairness,and fiber structure on the yarn surface,it is difficult to detect yarn diameter accurately and test results have a large discreteness.In this paper,firstly we find statistical laws between yarn diameter and hairness by a great deal of experimentation,and then we design recognition algorithms to remove effectively the effects of yarn hairness,so the yarn outline will be extracted successfully,finally we can get average of yarn diameter and unevenness index. On the basic of conversion law between yarn diameter and linear density of the yarn,we analyze reasons of deflection between ideal conversion laws and actual results,establish dynamic correction coefficient and realize automatic measurement of linear density of the yarn.Compared to the results of traditional methods,experimental results demonstrate that relative errors of image measurement lie mainly in 5%,which show the project has both practicable value and the theoretical value.To yarn twist angles examination,firstly it is thinned for the images which can weaken influence of yarn hairness,and then the mainly orientation of fibers on the surface of yarn is extract using Hough Transform to express yarn twist angles.Experimental results demonstrate that relative errors of image measurement lie mainly in 5%compared to the results of man-measurement,which shows that the system can realize detection with rapid and non destructive,and make up for detects of conventional methods.In short,the test system and computational methods provided in this paper change traditional test methods of fabric structural characteristics,relieve testing intensity,raise test efficiency and accuracy and has both practicable value and the theoretical value.At the same time,the equipment and algorithms we provided above have certain theoretical value and use for reference.