Yarn Hairiness Detection And Analysis Based On Image Technology

The number and length of yarn hairiness are directly determined the appearance, the quality and the grade of the yarn, and as well as the appearance of the fabric. The long hairiness of yarn is more, the opening of weaving is more difficult, and the dyeing is unevenness, which can cause the productivity of weaving and dyeing process are reduced. Therefore, hairiness detection is a key link in the textile product quality control. At present, domestic and foreign textile enterprises commonly used photoelectric grading statistical method and photoelectric full hairiness method. Because the limitations of the photoelectric and information collection, the grading statistical method cannot collect the complete hairiness information, the result is affected by test speed, and the reproducibility is poor. The detection results of full hairiness method cannot reproduce and the H-value is a unified estimate value which cannot visually reflect the yarn hairiness distribution, so the photoelectric method cannot meet the accurate monitoring requirements of yarn hairiness in modern textile production. The detection of yarn hairiness based on digital image acquisition and analysis technology is expected to overcome the shortcomings of the existing photoelectric method. However, the existing image method still cannot collect complete the yarn hairiness information in the past two decades, the main reason is that the image segmentation algorithm and the hairiness measurement algorithm lack the adaptability to the yarn hairiness which is complex and changeable in the space and disorderly in the extending direction.The aim of this research is to get the image segmentation algorithm and the hairiness analysis algorithm which can extract the complete hairiness information by the research of the yarn image processing process, detection and analysis algorithms of the hairiness skeleton(hairiness center line)and length. The influence of hairiness detectable angle on the measurement results was microscopically analyzed and the hairiness detectable rates of photoelectric method were obtained by establishing the hairiness detection model of photoelectric method. For the deficiency of the existing photoelectric method and image method, the fixed-length segmentation algorithm and fine-segmented algorithm of hairiness measurement based on hairiness skeleton were proposed to achieve the grading statistics of hairiness and fine measurement of single long hairiness and single short hairiness. In order to obtain a more adaptive and accurate measurement analysis algorithm of hairiness skeleton, the tracking algorithm of hairiness skeleton and length was introduced, the hairiness path matrix was generated by judging, confirming and recording the hairiness point, which can achieve tracking measurement of various hairiness. The implementation results showed that the hairiness skeleton tracking algorithm has better applicability than fixed length segmentation and hairiness segmentation, and can obtain more complete information of hairiness and more accurate hairiness length. The main contents and conclusions of this paper are as follows:(1) The yarn hairiness detection model was established, the influence of yarn diameter, hairiness length and hairiness relative angle on hairiness measurement and hairiness detection angle were analyzed, and the hairiness detectable rate was calculated to use correct the test result of photoelectric method. In order to make a detailed analysis of the factors that affect the hairiness measurement results and obtain the hairiness detectable rate of the photoelectric method, the hairiness testing model of the photoelectric method was constructed. Through the model analysis, the important role of hairiness length and hairiness relative angle in hairiness measurement and the detection of "blind area" in hairiness detection were determined. The detectable angle and detectable rate of hairiness were proposed to use as reference to measure the detection effect of photoelectric method. It not only provides a detailed analysis of the hairiness measurement from the microscopic level, but also provides guidance and direction for the image processing and hairiness analysis algorithms.(2) By analyzing the parameters and existing problems in every link of yarn image processing, the image processing steps of hairiness measurement were determined by contrasting the effects. In order to obtain a more complete yarn image, an image processing flow suitable for hairiness measurement is constructed. Firstly, the collected color original yarn image was converted into grayscale yarn image, and the gray distribution in the gray matrix of the image was analyzed in detail. By comparing the segmentation results of Otsu threshold method, dynamic threshold and manual threshold of yarn image. The adaptability of this series of yarn images acquired by MOTIC microscope is good, when the dynamic threshold is taken T3(The mean of the gray-scale peak of the histogram and the mean of the minimum gray-scale and the Otsu threshold) and the manual threshold is taken as 65.7(The secondary correction threshold of peak gray scale range from 50 to 75 in the gray level histogram). The disk structure operator with a radius of 11 was used in opening operation of binary yarn image and the yarn core can be quickly separated from the yarn binary image, the shape of the yarn core is good and yarn hairiness is complete. Zhang fast parallel thinning algorithm in Matlab was used to obtain the hairiness skeleton with good connectivity and good shape retention property can be used for the measurement of hairiness of various shapes.(3) The fixed-length segmentation algorithm and the fine-segmentation algorithm based on the hairiness skeleton are constructed and the hairiness skeleton is analyzed. The graded statistical results and accurate measurement results of yarn hairiness length are obtained.The fixed-length segmentation algorithm and the fine-segmentation algorithm are designed for segmentation and measurement of the hairiness skeleton and the length, which can achieve the accurate measurement of hairiness that cannot achieve in the existing image method. In the fixed-length segmentation algorithm, the edge curves of the yarn core were taken as the segmentation line, and the step value is 1mm. In the direction perpendicular to the yarn core, many segmentation lines paralleled to the yarn core were generated to segment hairiness skeletons. And the intersection points of the segmentation line and the hairiness skeleton were taken as the judgment objects of the hairiness length to count the hairiness number of different lengths. In the fine segmentation algorithm, the hairiness skeleton was subdivided into a number of small hairiness fragments with different initial segmentation lines and segmentation steps. The length of all the hairiness fragments is the whole length of hairiness. The experiment of hairiness measurement showed that the fixed-length segmentation algorithm can achieve statistical classification of hairiness and result is close to the result of artificial visual, which has bigger detectable angle range. The fine segmentation algorithm can achieve the fine measurement of hairiness, and can obtain more complete hairiness information then fixed-length segmentation.(4) The tracking algorithm of hairiness skeleton and the length was used to analyze and locate each hairiness point in a single hairiness skeleton, and a hairiness path matrix was obtained and used to calculate more accurate hairiness length. The fine-segmentation algorithm of hairiness skeleton and the length cannot extract specific information of hairiness fragment in the extraction of yarn feather skeleton information. Aiming at this problem, the tracking algorithm of hairiness skeleton and the length was designed to achieve the location of every the hairiness point on the hairiness the skeleton. The hairiness starting point and hairiness points were sequentially recorded into the hairiness path matrix, and the position information of every hairiness point and information of hairiness length were completely extracted. The length of hairiness was calculated by judging the relative position of two adjacent hairiness points. The experiments of tracking measurement of hairiness skeleton and the length showed that algorithm has certain adaptability to tracking measurement of single long hairiness, single short hairiness, single curly hairiness, overlapping hairiness, crossed hairiness and interrupted hairiness, and measured hairiness length is more accurate and more realistic.(5) The continuous yarn images were collected by MOTIC video microscope, the hairiness area index, the hairiness length index and the hairiness length and root number of different yarns were measurement and the hairiness of the four yarns were analyzed, the hairiness length and number index can be used to describe the specific distribution of yarn hairiness. In order to obtain the yarn image with less interference noise, the collection parameters were set to collect continuous yarn images of 14.6tex, 11.7tex, 9.7tex and 7.3tex of pure cotton combed yarn. The hairiness area index, hairiness length index and hairiness length and root number of the four yarns showed that hairiness area index can reflect the distribution of hairiness of the yarn itself but cannot describe the hairiness distributions of different yarns. Hairiness length index can be compared the distributions of different yarn hairiness, but cannot reflect the specific distribution of hairiness because of it is a non-dimensional units. The hairiness length and root number not only can reflect the specific distribution of the hairiness of the yarn itself, but also can be compared with the specific distributions of different linear density yarns.