Research Of Dynamic Identification Technology For Cotton Foreign Fibers

The project of Research of Dynamic Identification Technology For Cotton Foreign Fibers is supported by the National Science and Technology Pillar Program in the Eleventh Five-year Plan Period"The research and development of cotton processing complete set technical equip". (2006BAD11A14-3).Though the proportion of foreign fibers in cotton is little, it does great harm to the cotton; it would have the adverse effects to the quality of spinning. Thus affected the economic benefits of cotton textile industry, reduced the international competitiveness of domestic cotton. Dynamic identification technology for cotton foreign fiber to improve the technical level of the cotton quality inspection and to promote cotton fair deal has the very important significance.On the question of manual selection method has low efficiency and large error practical problems during the cotton foreign fiber content testing, this paper studied a method of dynamic identification technology for cotton foreign fiber based on machine vision system and realized the classification and weight statistics of cotton foreign fiber. The mainly following work has been completed:1. The identification system structure of cotton foreign fiber was designed; this system consists of image collection system and image processing system. Image collection system is responsible to collect the multi-spectrum dynamic images of cotton; image processing system is responsible to process the images in order to realize the classification and weight statistics of cotton foreign fiber.2. A kind of method of image segmentation was proposed which based on the Mean-shift adaptive threshold value segmentation technology; it isn't very high for the whole image of light and cotton layer background and it has very high robustness, also it can complete the image segmentation rapidly, can divide the target sector exactly and obtain the accurate foreign fibers image and so on.3. The image processing recognition plan of cotton foreign fibers was designed, first the original cotton foreign fiber image was carried on gradation processing in order to obtain the gray histogram. Through the analysis of gray histogram decided that uses the Mean-shift adaptive threshold value segmentation technology to complete the cotton foreign fiber image segmentation, then dilation process and median filtering process were used to complete morphology processing and enhancement processing of cotton foreign fiber in order to obtain clearly binary image, finally the hollowed inner point method was used to extract all the contours and 8-neighborhood search method was used to extract each object contour.4. The characteristic parameter of cotton foreign fiber contour image was extracted and was carried on data analysis based on rough sets, then the effective characteristic attribute values of cotton foreign fiber were extracted, Finally the work of cotton foreign fiber classification and identification was completed based on the decision tree method.5. A method of weight statistics of cotton foreign fiber was proposed, weight error test was carried and a model of weight statistics of cotton foreign fiber was established.This paper chooses 9 kinds of foreign fibers, such as the black plastic piece,the blue silk,the red silk,the hemp rope,the hair,the non-fluorescent red polypropylene silk,the fluorescent white plastic piece,the fluorescent white polypropylene silk and the feather as object of study, and was carried on image processing,classification and weight statistics. It's proved by experiments that the cotton layer detection rate of this method reached 40m/h, the identification precision achieved above 95% and the weight statistics error less than±8%.