Automatic Detection Of Cotton And Flax Fibers Based On Image Processing

The content of cotton and flax fibers in cotton and flax fabrics have a great influence on the price and performance of textiles.Therefore,it is necessary to detect the content of cotton and flax fibers in cotton and flax fabrics during production and sales.At present,the fiber inspection organization mainly uses the microscope recognition method for fiber recognition,but the efficiency of human eyes recognition is low and affected by human subjective.The combination of computer image processing technology and textile fiber recognition can effectively improve the recognition accuracy and work efficiency.In this thesis,the images of cotton / flax fiber are collected under the microscope.After preprocessing the images,the characteristic values of the fiber have extracted.After inputting the parameter of the characteristic values into BP neural network training,the automatic recognition of cotton fibers and flax fibers can be completed,and the blending ratio of the sample cotton / flax blended yarn is calculated.The main research contents are as follows:(1).In this thesis,in the fiber pretreatment stage,the effect of different algorithms is compared,and finally the pretreatment route with good effect and high efficiency is selected.In this thesis,the weighted average method is used for gray-scale processing,the median filter is used for noise removal,the fiber images are strengthened,the improved Otsu method is used for binarization,and the skew of fiber images are corrected.Finally,the improved Canny algorithm,morphological operation and the combination of edge scanning are used for the preprocessing route of contour extraction.(2).In this thesis,six characteristic values of cotton / flax fiber images are extracted,which are diameter ratio,diameter standard deviation,maximum twist,average twist,overall fullness and fullness standard deviation.Through the comparison and analysis of the eigenvalue parameters,the adaptive threshold is set,and the cotton / flax fiber types are identified automatically by the threshold judgment.Due to the large numerical span between different eigenvalue parameters,this thesis normalizes all the eigenvalue parameters,so that the range of the eigenvalue parameters are between [0,1].When the normalized eigenvalueparameters are used for subsequent calculation,the operation speed can be effectively improved.In this thesis,the correlation analysis of six eigenvalues is carried out.After calculation,the correlation coefficients between the six eigenvalue parameters and the output fiber types are all above 0.9.The result shows that the six eigenvalues are highly correlated with the fiber types.(3).In this thesis,the parameters of BP neural network are set,the input layer is 6 kinds of eigenvalue parameters,and the output layer is fiber type after the calculation of hidden layer.After BP neural network training,the recognition accuracy of cotton fiber image is95.1%,and that of flax fiber image is 93.8%.The experimental result shows that the recognition pattern based on the combination of 6 eigenvalues and BP neural network can effectively recognize cotton / flax fibers,with high accuracy and fast recognition speed.(4).Through the automatic detection system of cotton / flax fibers,the diameter of each sample fiber can be counted when the diameter characteristic value is extracted;after the detection and recognition,the number of cotton fibers and flax fibers can be counted automatically.The blending ratio of sample cotton / flax blended yarn can be calculated by the formula of blending ratio of cotton / flax yarn and the data.The blending ratio of cotton /flax blended yarn selected in this project is 70 / 30.After calculation by the cotton / flax fiber automatic detection system,the blending ratio of the sample yarn is 70.8 / 29.2,with an error of 0.8%,which conforms to the yarn blending ratio detection standard.