Identification Of Wool And Cashmere Based On Spectral-Line Analysis

Identification of wool and cashmere is one of the most challenging propositions in textile industry. Among the currently methods that have been proposed to distinguish cashmere from wool, the microscopy method keeps the most important and extensive place. But it is not easy for optical microscopy to obtain clear images of the scales, besides optical microscopy highly depends on the experience of the operator. The scanning microscopy can present the surface information of the scales clearly, but it requires heavy investment. Hence rapidly, objective and accurately recognition systems are impending in wool industry.This paper proposes a method to identify the spectral line of wool and cashmere to distinguish wool from cashmere. Through refining and clustering the surface texture of wool and cashmere, the texture gray images are obtained, followed by dilating and distance retracting the texture gray images, texture assembled images of wool and cashmere are generated. To analyze texture groups that represent the scale feature of wool and cashmere, the texture has been projected onto a designated axis to form a curve called spectral line.This paper randomly selects 330 wool and cashmere respectively to extrapolate the identification algorithm. The research starts with the segmentation of the spectral lines. Six features are extracted from the segmented spectral lines via MATLAB. Three features are screened out through analyzing the probability distribution curves of the six features. After correlation analysis of the three screened features, two features are selected to establish the identification function to identify wool and cashmere. Since there are intersections in the drop-point-graphs of the selected features of wool and cashmere, low-pass filtered processing is employed to reduce such intersection. The comparison of the drop-point-graphs of wool and cashmere illustrates low-pass filtered processing make positive contribution to the identification. Initial tests on 660 randomly selected samples indicate that the discrimination error is 1.2%. 2060 wool and cashmere blended samples are identified to further investigate the validity of the proposed algorithm. Final result shows that the identification rate can up to 96.63% based on this algorithm. Besides, the identifications of 70 ingenious wool fibers are evaluated. Although the algorithm makes the identification of ingenious wool fiber and cashmere easy, it still cannot get perfect result. But the huge difference of the coefficient of fineness variation between ingenious wool and cashmere fiber increase the identification rate. The stability of the algorithm is verified by analyzing the related degree, the sample size and the ratio of wool and cashmere in the blended samples.