Research On System Of Recognition And Content Automatic Analysis In-Situ Fiber In Textiles

China's cashmere production accounts for around 70% of world production, which is the world's largest cashmere processing and exporting country. Cashmere's yield is low. The world's annual cashmere production is about 15000-20000 tons, about 1% of total production of wool, and 0.2% of total production of animal hair fiber. And the price of Cashmere is very high. In 2009, domestic cashmere price is about 400-650 Yuan/kg, which is more than 10 times the price of wool. Also with characteristics of slender soft, rich luster and light thermal, cashmere becomes a high-value fiber, and cashmere products also become valuable items. Which also makes cashmere fibers easily be mixed with other fibers for phishing. Among that, fine wool is very similar with cashmere on morphological, physical and chemical properties, so phishing among them is more serious. At present, there's the national compulsory testing standard (GB 18267-2000) already, which is effectively in identifying cashmere and measuring cashmere content. But it base solely on artificial discrimination, so it is not only time-consuming and low accuracy. Therefore, the only way to resolve this hard problem is to research an automatic identification method and formulate a sound appraisal standard for finished cashmere and wool products.With development for years, there are already a dozen methods for cashmere and wool identification related to chemistry, biology, image processing and other areas. Of this technical means only infrared spectroscopy, gene-chip method, image processing method can automate. Because of poor practicality, poor reliability, complex sample preparation, system's complexity and its operational complexity, the first two methods with the principle of component identification are difficult to achieve full automation and instrumentation. Obtain images of cashmere and fine wool through the optical fiber microscopy and scanning electron microscopy, then extract characteristic parameters from the image by image processing methods, at last cashmere and wool can be distinguished automatically in use of classification models. Only optical microscope can get non-destructive sample images with low cost fast through simple operation. And it is easy to spread commercially.The Sample unit is formed of pressurized devices, sample tank and the glass cover, which limit the fibers on surface of samples into objective depth without causing damage. Then a Len with a long working distance can be used to collect images of the fibers on the inner surface of the glass.Because of high precision and less vibration of the instrument hardware and motion mechanism, when the sample unit is running, samples move up and down little. Therefore, the Len can be focused negatively. As one method of Automated Optical Inspection, the Read-on-the-fly is chose to get image fast in real-time. When CCD captures images, the sample unit capture images automatically along 's' type path. Then the instrument can capture images of the fibers in-situ, rapidly, continuously, non-destructive and real-time.Digital optical system consists of a Len of a long working distance, large depth, high-magnification, a high exposure speed CCD and a reflective lighting device. Sample pressure, oil-soaked, dark-field illumination, glares eliminating and other channels has improved the image quality significantly. Lighting installations should be large-aperture, high-brightness, long-distance, and smaller angle of the projection, dark-field effect can obtained from which when objective lens at a high NA and can meet the requirements of brightness and brightness uniformity for objective.The Instrument's image acquisition rate is about 300-600 pieces / minute, and the sampling rate of it is about 200-400 roots / min. Image processing method is used to extracting the characteristic parameters of multi-fibers in the whole image, which improves the processing speed. The Bayesian classification model is used to identify two kinds of fibers, and results show that the actual identification accuracy is about 87.5%. The system is feasible, and basically achieved features of non-destructive, rapid and accurate detection. That is a successful attempt for the development of mature and reliable utility-type equipment and has laid a certain foundation for that.