Detection Of Yarn Evenness By Means Of Digital Image Processing

In order to overcome the deficiencies and shortcomings of capacitive evenness testerwhich has harsh demanding to test environment and its capacitance mouth has average effect,This paper presented a detection method of yarn evenness based on digital image processing.In the experiments, the MRS-4800M48U scanner and Motic SMZ-140type of videomicroscopy were used to capture yarn images. Then, the yarn images were treatedsequentially with filtering pretreatment, binarization and morphology opening operation. Theyarn diameters、the CV values and thin places of yarns can be calculated according to thetreated images. Detect the yarn evenness by comparing variation of yarn diameters, yarn CVvalues of the different fragment length and yarn thin distribution. Throughout the process ofyarn image acquisition, image processing of yarn and yarn evenness testing, this paper has thefollowing main conclusions:(1) When capture yarn images with scanners and video microscopy, high resolution andhigh magnification should be selected in order to make the captured images clear enough toachieve the identification requirements. This paper selects the resolution of the scanner is3810dpi and40times magnification was selected for the video microscope after severalcomparision experiments. Existed light source of video microscopy can make image qualitymare stable compared to selecting natural light and fluorescent light as light source of capture.When adjust the focal length of the video microscope, set the yarn15mm higher than thebackground, which can highlight the yarn, dilute the background and make captured imagesprocessed more easily.(2) Quality of yarn image processing affects the accuracy of following yarn evennessdetection directly. In allusion to the problem of noise and slender yarn hairiness in yarnimages, this paper pretreated the images with two-dimensional adaptive Wiener filteringfirstly to remove the noise and slender yarn hairiness. It was shown by comparison thatchoosing the filter window size of45pixels×45pixels and65pixels×65pixels can gainbest results for yarn images captured by scanners and video microscopy. The treated yarnimages almost have been no noise or elongated feathers. In order to increase the contrast ofyarn trunk and background, extract the yarn trunk, yarn the images were segmented by Otsuthreshold segmentation in this article. Segmentation results show the Otsu thresholdsegmentation method can separate the yarn and background, which made the yarn edgesclearer except some burrs and small points; In this article, morphological opening operationwere used to process the segmented yarn images to eliminate burrs and small points in theimages. On trail, disc structure elements with radius6and10can gain best results inprocessing the binary yarn images(3) Yarn diameter value detected with digital image processing method proposed in thispaper is very close to the theoretical value, the maximum error was less than3%; The longerthe of the yarn CV value detected had high accuracy, choosing the longer test fragment, thesmaller the CV values would be. Compared detected the yarn thin place results and yarndiameter deviation spectrogram, we found the thin place and troughs of the wave spectrum position were corresponding.(4) In order to provide a friendly interaction, this paper designed yarn evenness testinginterface system according to yarn image processing and testing process of yarn evennessdetection. People can detect the yarn evenness by the interface system more conveniently andintuitively.