Identification Of Cashmere And Wool Based On Texture Analysis

Identification of cashmere and wool is an important research subject in textile industry. There are many identification methods now, from which optical microscope(OM) and scanning electron microscope(SEM) are more recognized by AATCC and IWTO. SEM is a relatively accurate method to separate cashmere from wool, but more time and cost will be taken. Moreover, clear scale images of cashmere and wool are hard to obtain and the accuracy is depended on human’s experience. Therefore, in order to decrease labor intensity, it’s urgent to explore a digital recognition system that can improve identification accuracy and efficiency.The method used in this paper is texture analysis. The texture of scale is global image feature. Texture is the physical structure information which is implied in the difference of image gray level, and it is an embodiment of the law of gray level change. In this paper, one of the most important texture analysis methods is gray level co-occurrence matrix(GLCM), which will be used here to describe the regularity of gray level distribution. GLCM method is used to describe the texture characteristics of the image by the continuous change of the gray level of the adjacent pixels in the image.First of all, through exploring the influence of distance d, image gray level k and scanning direction θ for the generation of the gray co-occurrence matrix, the method suitable for identification of cashmere / wool is established based on their own characteristics. The result is that d = 5, k = 256, θ will be 0°, 45°, 90°, 135° and calculating the feature parameters from each direction and get the average of feature parameters from four directions. Finally, 14 texture features of cashmere / wool’s scales are extracted. Due to the 14 texture features are not totally linearly independent; the method of principal component analysis(PCA) is used to reduce data dimensions and to get independent principal texture features that can describe texture features of cashmere / wool better and efficiently. Finally, building BP neural network model and the model will be repeatedly trained to become the best network structure. At last we get a high identification accuracy for cashmere / wool. The results of the research show that there is a large difference between the texture characteristics of cashmere and wool, the use of texture analysis method to identify the cashmere wool has a high degree of accuracy.