Detection Of The Blending Ratio Of Cotton And Kapok In Nonwoven Webs Based On Image Processing

Identification of textile fiber is an important part of China's textile inspection. At present, the commodity inspection authorities mainly adopt artificial recognition method to implement determination of the blending ratio of kapok and cotton, the accuracy and efficiency of this approach mainly depends on the experience and technique of testing personnel, so the efficiency is very low with poor stability. In recent years, the application of kapok has become increasingly widespread in nonwoven material because of its excellent characteristics. So we need an efficient and accurate inspection method for the identification of cotton and kapok. This paper uses image processing technology to achieve the determination of blending ratio of kapok and cotton in nonwoven webs.During the process of lapping kapok/cotton nonwoven webs, we found out common blending ratios in production by blending kapok and cotton of different proportion and combining with the difficulty of lapping webs. In the process of sample-making, the suitable length of the fibers in longitudinal segments was determined by the contrast tests. Use full automatic fiber fineness instrument for image acquisition, and then pre-process fiber images, including image segmentation, morphological processing, removal of noise points, filling holes and other processes.Because there are a large number of overlapped fibers in the vertical production, so it is necessary to separate overlapped fibers.This papper divides the common intersection conditions into four kinds, including decussation, T cross, end to end and paralleled fibers. The paralleled fibers can be separated according to the concavity and convexity of the external contour. The intersection conditions of decussating, T cross, end to end can be effectively separated according to skeletons' trend, decussation and T cross can achieve defibration because the infall slope difference of same skeleton is smaller. In the case of end to end, we separate fibers at the intersection according to the characteristics of mutations in the adhesion of skeleton slope point. Through the above treatments, the overlapped fibers are separated into independent targets. According to gray level difference of the fibers, we select two parameters of local high aspect ratio, gray variance of each histogram as effective characteristic parameters to discriminate cotton and kapok. Classify the targets and build cotton and kapok formula recognition according to the classic pattern recognition method.During the process of calculating blending ratio, because there is deviation of measured width and the real diameter, we firstly finish the conversion from projected diameter of target fiber to real diameter, and then calculate the blending ratio of kapok and cotton in nonwoven webs.After testing, for the high quality of the chip, the accuracy of measured blending ratio can reach more than 90%.