The Research On Yarn Hairiness Detection Algorithm Based On Image Processing

Textile develops in the direction of high quality and high value-added on the condition that high quality yarn is produced.Therefore,the detection and evaluation of yarn quality becomes a crucial problem to improve the quality of textile.Yarn hairiness is an important parameter in many parameters of yarn quality detection,which not only has a great impact on appearance,handing of yarn and final textile as well as back-end processing,but also pollutes the environment and harms health.At present,the hairiness detection methods have been launched such as visual observation method,weighing method and photoelectric method,which exist some disadvantages in detection precision,detection efficiency and cost.With the development of computers and image processing technologies,the application of computer in the field of yarn quality control have drawn more and more attentions.It is of great significance to achieve objective assessment and improve yarn quality by making full use of image processing technologies to overcome the drawbacks of other detection methods.This paper mainly researches on yarn hairiness detection algorithm based on image processing,which combined with image acquisition device.The algorithm can be divided into six aspects of image preprocessing,image segmentation of yarn hairiness,yarn hairiness extraction,yarn hairiness statistics,the overall analysis of algorithm and the production of Graphical User Interface.According to the initial level of the captured images,two methods are proposed in image preprocessing,which are gray-scale transformation and skew correction.Image segmentation of yarn hairiness is the basis of the subsequent algorithms,whose result will have direct effect on the accuracy of overall algorithm.It contains two sections: yarn segmentation and yarn stem segmentation.Yarn is segmented by using a multi-resolution Markov Random Field model with a variable weight in the wavelet domain,and yarn stem is segmented by parametric kernel graph cut method.Yarn hairiness extraction consists of hairiness binary-image extraction and hairiness thinning.Two calculation methods of hairiness length are proposed and comparatively analyzed in the step of yarn hairiness statistics on the premise of unit calibration,which are statistical lines method and actual length tracking method.The feasibility analysis of overall algorithm is carried in the standard of detection results from visual observation method and USTER instrument respectively.Experimental results show that,compared with visual observation method,the maximum deviation ratio of proposed overall algorithm is 6.58%.Meanwhile,the accuracy rate of proposed overall algorithm reaches up to 95.5%,compared with the other two methods in the standard of test results of USTER CLASSIMAT 5.It is proved that the proposed yarn hairiness detection strategy is feasible,and high robustness processing procedure,straightforward results,controlled detection precision and intelligent detection program are realized,which provides an important referenced value for building the better-developed yarn hairiness detection system and assessment system.Finally,a friendly Graphical User Interface is designed according to the proposed overall algorithm,the completed researches are summarized,and the future trend of this subject is forecasted.