Research On Image Identification Of Cashmere And Wool Fiber

The identification of cashmere and wool has always been an important topic in the textile field.At present,the commodity inspection department mainly adopts the micro-projection method,which has low detection efficiency and is greatly influenced by man-made subjective.With the rapid development of computer technology,the method of combining microscope with computer to detect cashmere and wool is becoming more and more mature.In this paper,Based on the image processing technology,the image preprocessing method,cross fiber separation,feature extraction and classification recognition model of cashmere and wool are analyzed.And an automatic recognition and classification method is realized.The main research contents are as follows:Firstly,an image preprocessing scheme for cashmere wool fiber is determined.Using automatic fiber slicing instrument and differential interference contrast microscope to collect images with clear texture features.According to the characteristics of cashmere and wool fiber image under microscope,a series of image processing methods such as graying,Laplacian sharpening,Gaussian filtering,Canny detecting,morphological dilating and eroding,contour detecting and region growing method are adopted.The target fiber and background are segmented,and the binary graph of the target fiber region is obtained.Secondly,a cross fiber separation algorithm is proposed.Aiming at the binary graph of the target fiber region,the improved fast parallel thinning algorithm is used to obtain the skeleton graph.Using the number of crossing points of the skeleton as the classified criterion,the fiber morphology is divided into five types: single fiber,X shape,T shape,V shape and multiple cross fibers.For the cross shape of two fibers,After being processed by Shi-Tomasi corner detection and Hough transform,they are separated into two single fibers according to the principle of the similar slope of the same root fiber.For multiple cross fibers,they are identified by human-computer interaction.Thirdly,the morphological feature and texture feature extraction method of cashmere wool fiber are studied.For single fiber,eight morphological features,such as fiber diameter,scale height,scale perimeter,scale area,ratio of diameter to height,scale relative perimeter,scale relative area,density and so on,are extracted by means of piecewise measurement,scale skeleton measurement and statistical marking point method.And four texture features,such as energy,entropy,contrast,correlation and so on,are extracted based on the gray level co-occurrence matrix.These twelve characteristic parameters can describe the fiber characteristics more comprehensively and have better separability.Finally,a classification identification model of cashmere wool based on principal component analysis and support vector machine is established.The dimension reduction of morphological and texture feature data is carried out by principal component analysis,and the recognition parameter system of cashmere wool is established.And then selecting RBF kernel function to obtain the optimal parameter combination through the cross-checking method and the grid search method,and training to obtain a support vector machine classification model which can be applied to the cashmere wool fiber identification.The experimental results show that the classification model is effective and the recognition rate is up to 94.9%.